Swearing in English (Routledge Advances in Corpus Linguistics)

Citation preview

Swearing in English Swearing in English uses the spoken section of the British National Corpus to establish how swearing is used, and to explore the associations between bad language and gender, social class and age. The book goes on to consider why bad language is a major locus of variation in English and investigates the historical origins of modern attitudes to bad language. The effects that centuries of censorious attitudes to swearing have had on bad language are examined, as are the social processes that have brought about the associations between swearing and a number of sociolinguistic variables. Drawing on a variety of methodologies, including historical research and corpus linguistics, and a range of data such as corpora, dramatic texts, early modern newsbooks and television programmes, Tony McEnery takes a sociohistorical approach to discourses about bad language in English. Moral panic theory and Bourdieu’s theory of distinction are also utilised to show how attitudes to bad language have been established over time by groups seeking to use an absence of swearing in their speech as a token of moral, economic and political power. This book provides an explanation, not simply a description, of how modern attitudes to bad language have come about. Tony McEnery is Professor of English Language and Linguistics at Lancaster University, UK, and has published widely in the area of corpus linguistics.

Routledge advances in corpus linguistics Edited by Tony McEnery Lancaster University, UK and Michael Hoey Liverpool University, UK

Corpus-based linguistics is a dynamic area of linguistic research. The series aims to reflect the diversity of approaches to the subject, and thus to provide a forum for debate and detailed discussion of the various ways of building, exploiting and theorising about the use of corpora in language studies. 1 Swearing in English Bad language, purity and power from 1586 to the present Tony McEnery 2 Antonymy A corpus-based perspective Steven Jones 3 Modelling Variation in Spoken and Written English David Y.W.Lee 4 The Linguistics of Political Argument The spin-doctor and the wolf-pack at the White House Alan Partington 5 Corpus Stylistics Speech, writing and thought presentation in a corpus of English writing Elena Semino and Mick Short

6 Discourse Markers Across Languages A contrastive study of second-level discourse markers in native and non-native text with implications for general and pedagogic lexicography Dirk Siepmann 7 Grammaticalization and English Complex Prepositions A corpus-based study Sebastian Hoffman 8 Public Discourses of Gay Men Paul Baker

Swearing in English Bad language, purity and power from 1586 to the present

Tony McEnery

LONDON AND NEW YORK

First published 2006 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 270 Madison Ave, New York, NY 10016 Routledge is an imprint of the Taylor & Francis Group This edition published in the Taylor & Francis e-Library, 2005. To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to http://www.ebookstore.tandf.co.uk/. © 2006 Tony McEnery All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A catalog record for this book has been requested ISBN 0-203-50144-6 Master e-book ISBN

ISBN 0-203-59882-2 (OEB Format) ISBN 0-415-25837-5 (Print Edition)

This book is dedicated to those who struggle to have their views heard

Contents List of figures

vii

List of tables

x

Acknowledgements

1 Bad language, bad manners PART 1 How Brits swear 2 ‘So you recorded swearing’: bad language in present-day English PART 2 Censors, zealots and four-letter assaults on authority

xiv

1 23

24 51

3 Early modern censorship of bad language

52

4 Modern attitudes to bad language form: the reformation of manners

71

5 Late-twentieth-century bad language: the moral majority and four-letter 102 assaults on authority PART 3 Discourses of panic

130

6 Sea change: the Society for the Reformation of Manners and moral panics 131 about bad language 7 Mutations: the National Viewers’ and Listeners’ Association moral panic 166

Postscript

204

Notes

207

Bibliography

236

Index

243

Figures 1.1

A letter appearing in the autumn 1999 issue of the National Viewer and Listener

7

1.2

A sample collocational network

19

1.3

The network around swearers in the SRMC

23

2.1

Frequency of BLWs per million words in groups of different ages

39

2.2

Frequency of BLWs per million words of speech produced by different social classes

42

5.1

The linguistic mandate of power

112

5.2

An excerpt from Till Death Us Do Part, transmitted 11 October 114 1972 (‘Dock Pilferring’)

5.3

An excerpt from Steptoe and Son, broadcast 27 March 1972 (‘Divided We Stand’)

118

6.1

Four examples of the consequences of guilt

137

6.2

Four examples of the consequences of wrongdoing for the public

138

6.3

Four examples of the nature of the judgement which will be brought on those guilty of sin

139

6.4

Concordances of ourselves

142

6.5

A sample concordance of swearing

143

6.6

Four examples of men meaning males

144

6.7

The discourse of moral panic in action

146

6.8

Three examples of the use of etc.

149

6.9

Will in passive constructions

149

6.10 A directional graph of the collocates of swearing

155

6.11 A directional graph of the collocates of drunkenness

155

6.12 Two graphs joined to form a network

157

6.13 Objects of offence and their linking collocates

158

6.14 Collocates of common

160

6.15 Common meaning something shared by all

161

6.16 Common meaning something that is usual

161

7.1

Words which are key-keywords in five or more chapters of the MWC

170

7.2

Words which are key-keywords in all of the MWC texts

170

7.3

The responsible

176

7.4

Porn is good

184

7.5

The call for the restoration of decency

189

7.6

Pronoun use by the VALA

190

7.7

The assumption of Christianity

190

7.8

Speaking up for the silent majority

191

7.9

The use of wh-interogatives by the VALA

192

7.10 The major collocational network in the ‘permissive society’ grouping

194

7.11 Four-letter assaults on authority

201

Tables 1.1

Text categories in the Brown corpus

13

2.1

The categorisation of bad language

32

2.2

Categories of annotation

27

2.3

Words preferred by males and females in the BNC ranked by LL value

29

2.4

A scale of offence

30

2.5

Table 2.3 revisited—BLWs typical of males and females mapped onto the scale of offence

31

2.6

Categories of BLW use more typical of males and females ranked by LL value

31

2.7

Patterns of male/female-directed BLW use

33

2.8

Words more likely to be directed by females at either males or females ranked by LL score

33

2.9

Words more likely to be directed by males at either males or females ranked by LL score

34

2.10 BLWs directed solely at males and females ranked by frequency of usage

35

2.11 Table 2.8 revisited—BLWs typical of females used either of males or females mapped onto the scale of offence

36

2.12 Table 2.9 revisited—BLWs typical of males used either of males or females mapped onto the scale of offence

36

2.13 Average strength of BLWs in each category

37

41 2.14 The most frequent and least frequent users of particular BLW categories, categories ranked by strength from highest to lowest 2.15 The top-four BLW categories for each age group

41

2.16 The number of words spoken by three categories of speaker in the spoken BNC

45

2.17 The interaction of age and sex, frequencies given as normalised 47 counts per million words 2.18 The number of different word forms used to realise BLW use by the different age groups in the LCA

48

2.19 The distribution of three BLWs by age and social class, frequencies given as normalised counts per million words

49

4.1

Local and regional Societies for the Reformation of Manners in 80 England in the early eighteenth century

4.2

The expansion of the distribution of propaganda by the SRM, 1725–1738

84

4.3

Prosecutions for swearing and cursing brought by the SRM

91

5.1

The uses of bad language in Steptoe and Son (‘Men of Letters’) 120 and Till Death Us Do Part (‘The Bird Fancier’)

6.1

Positive and negative keywords in the SRMC when compared to the Lampeter corpus

132

6.2

A comparison of the SRMC and Lampeter B, yielding keywords for the SRMC texts

132

6.3

A comparison of Lampeter A and B, yielding keywords for the religious texts

133

6.4

A comparison of the SRMC with Lampeter A, yielding

134

keywords for the SRMC 6.5

The positive keywords of the SRMC/Lampeter B comparison categorised according to the major themes of a moral panic discourse

136

6.6

Consequence keywords

137

6.7

Corrective action keywords

139

6.8

Desired outcome keyword

141

6.9

Moral entrepreneur keyword

141

6.10 Object of offence keywords

142

6.11 Scapegoat keywords

144

6.12 Moral panic rhetoric keywords

147

6.13 Coordination of objects of offence in the SRMC

151

6.14 Words coordinated with keywords in the SRMC

152

6.15 Convergence in the moral panic

153

7.1

Keywords of the MWC when compared with the LOB corpus

167

7.2

Keywords in the MWC derived from a comparison of FLOB

167

7.3

The keywords of the MWC placed into moral panic discourse categories

168

7.4

Words which are key-keywords in five or more chapters of the MWC mapped into the moral panic discourse roles

170

7.5

Words which are key-keywords in all of the MWC texts mapped into their moral panic discourse roles

171

7.6

The distribution of chapter only, text only and chapter and text key-keywords across the moral panic discourse categories

171

7.7

The key-keyword populated model

173

7.8

Consequence keywords

174

7.9

Corrective action keywords

176

7.10 The keyword report

179

7.11 Collocates of pornography and pornographic

183

7.12 Enclitics which are negative keywords in the MWC when the MWC is compared to the sub-sections of LOB

184

7.13 The relative frequency of genitive’s forms and enclitic’s forms in the MWC compared to the sub-section of LOB

185

7.14 The collocates of programme and programmes

187

7.15 The collocates of film

187

7.16 The collocates of television and broadcasting

188

7.17 The collocates of decency

188

7.18 Object of offence keywords

193

7.19 The most frequently coordinated nouns in LOB

195

7.20 The most frequently coordinated nouns in the MWC

195

7.21 Top-ten key semantic fields in the MWC

198

Acknowledgements I cannot think of anything that I have ever written which owes so much to the comment and insight of others. I have spent the past eight years, on and off, talking about the ideas in this book to a range of researchers. Because of the nature of this book, the researchers I have spoken to have spanned a range of disciplines. I have also had many audiences, some shocked, some reflective, listen to and comment on the ideas presented here. While the list of people I would like to thank is enormous, I will limit myself here to people who have either suffered my musings on this topic at length, or whose contribution to this work, whether they know it or not, has been significant. From Linguistics at Lancaster I would like to thank Paul Baker, Norman Fairclough, Costas Gabrielatos, Andrew Hardie, Willem Hollman, John Heywood, Geoff Leech, Mark Sebba, Jane Sunderland, Andrew Wilson, Ruth Wodak and Richard Xiao in particular. Four other Lancastrians who deserve a mention are a historian, Michael Seymour, two of my colleagues from Religious Studies, Ian Reader and Linda Woodhead, and Paul Rayson from Computing. All four read parts of this book and gave me very useful comments from the perspective of their own disciplines. Beyond Lancaster, I would like to thank the following academics: Mike Barlow, Lou Burnard, Ron Carter, Angela Hahn, Mike Hoey, John Kirk, Merja Kyto, John Lavagnino, Barbara Lewandowska, Willard McCarty, Ruslan Mitkov, Geoffrey Sampson, Mike Scott, Harold Short, Joan Swann and Irma Taavitsainen. I also need to thank Matthew Davies who assisted me with library research for this book and Dan MacIntyre who helped to construct the corpora used in Chapter 7. On an institutional level, I would like to thank the Faculty of Social Sciences at Lancaster who provided me with two small grants to construct some of the corpora used in this book and a third grant to conduct work at the British Library. I must also thank the British Academy which has funded my work on seventeenth-century newsbooks as used in part in Chapter 3. The Libraries of Cambridge and Oxford Universities, as well as the British Library, were of enormous assistance in the writing of this book, particularly Chapters 3 and 4. I am indebted to them for their willingness to help. Additionally, Lancaster University Library, by giving me access to both its rare books archive and Early English Books Online, made my work much, much easier.

1 Bad language, bad manners Bad language Consider the word shit. Simply being asked to do this may have shocked you. Even if it did not, most speakers of British English would agree that this is a word to be used with caution. Because of prevailing attitudes amongst speakers of the English language, using the word may lead any hearer to make a number of inferences about you. They may infer something about your emotional state, your social class or your religious beliefs, for example. They may even infer something about your educational achievements. All of these inferences flow from a fairly innocuous four-letter word. Shit, and all other words that we may label as bad ‘language’, are innocuous in the sense that nothing particularly distinguishes them as words. They are not peculiarly lengthy. They are not peculiarly short. The phonology of the words is unremarkable. While it might be tempting to assume that swear words are linked to ‘guttural’ or some other set of sounds we may in some way impressionistically label as ‘unpleasant’, the fact of the matter is that the sounds in a word such as shit seem no more unusual, and combine together in ways no more interesting, than those in shot, ship or sit.1 A study of bad language would be relatively straightforward if this were not the case. So how is it that such an innocuous word is generally anything but innocuous when used in everyday conversation? How is it that such words have powerful effects on hearers and readers such as those you may have experienced when you read the word shit in the first sentence of this book? The use of bad language is a complex social phenomenon. As such, any investigation of it must draw on a very wide range of evidence in order to begin to explain both the source of the undoubted power of bad language and the processes whereby inferences are drawn about speakers using it. The potent effects of words such as shit can only be explained by an exploration of the forces brought to bear on bad language in English through the ages. It is in the process of the development of these attitudes that we see taboo language begin to gain its power through a process of stigmatisation. This process leads a society to a point where inferences about the users of bad language are commonplace. The following chapters will aim to add weight to this observation. For the moment, the reader must take this hypothesis on trust, as before we can begin the process of outlining evidence to support this hypothesis, a refinement of the goals of this book, and some basic matters relating to the sources of evidence I will use, need to be dealt with. The focus of this book is bad language in English, with a specific emphasis on the study of swearing. Bad language, for the purposes of this book, means any word or phrase which, when used in what one might call polite conversation, is likely to cause

Swearing in English

2

offence. Swearing is one example of bad language, yet blasphemous, homophobic, racist and sexist language may also cause offence in modern England. However, this book will not study changes in what has constituted bad language over the centuries. Books such as Montagu’s (1973) Anatomy of Swearing and Hughes’ (1998) Swearing have explored these changes already. Nor will this book work through a history of the changing pattern of usage of swear words as Hughes and Montagu have. Rather, this book has three distinct goals. First, it will study the effect of centuries of censorious attitudes to bad language. Following from this, this book will explore how bad language came to be viewed as being associated with a range of factors such as age, education, sex and social class. The passing parade of words that constitute bad language seems to have had little or no effect on what is associated with the users of bad language over the past three centuries or so. This book aims to look beyond the words that have caused offence to look for the social processes that have brought about the associations between bad language and a number of sociolinguistic variables. Finally, this book will seek to demonstrate that the roots of modern English attitudes towards bad language lie in the late seventeenth and early eighteenth centuries. It is in this period that we can find a social and moral revolution occurring which defined attitudes to bad language for centuries to come and established a discourse of purity as a discourse of power. In pursuit of the later two goals, this book explores the ways in which the public perception of bad language over the past 400 years has changed. The review is not comprehensive in the sense that I do not slavishly work through each decade and century. Rather I seek, by a study of three periods (1586–1690, 1690–1745 and 1960–1980), to outline the role that bad language has played in public life and public discourse in England. In doing so, I will investigate how the state has used bad language as an excuse for censorship (1586–1690), how bad language became associated with a number of sociolinguistic variables such as age, sex and social class (1690–1745), and how a discourse of power based on the absence of bad language was reinforced and defended in the debate over bad language in the media (1960–1980). In looking at these three periods, I will also argue that the studies presented are cumulative—in the later period the discourse of purity that was being defended was that established in the period 1690–1745, and in turn that linguistic purity was used as a tool of censorship in a way just as effective as any act of state censorship in the period 1586–1690. The goals link to the organisation of this book. The book is split into three major parts. In the first part, I pursue the first goal of the book by looking at the way in which modern English reflects historical processes which have formed attitudes to bad language. In the second part of the book, I will explore in detail what these historical processes were and how those processes have linked bad language to the demographic variables studied in Part 1. In exploring these historical processes I will look at both the establishment of these attitudes (1690–1745) and a recent example of the maintenance of these attitudes (1960–1980). In the final part of the book, I will look at the discourses which were used to establish and to maintain these attitudes. These three sections support a number of claims about bad language in modern British English. I summarise these claims here, though for the moment I will not seek to justify them—that is the work of the rest of this book. My claims are: 1 modern attitudes to bad language were established by the moral reform movements of the late seventeenth and early eighteenth centuries;

Bad language, bad manners

3

2 these attitudes were established to form a discourse of power for the growing middle classes in Britain; 3 the moral and political framework supported by a discourse of power can be threatened by the subversion of that discourse. In pursuit of my goals, I will need to use a wide range of sources of data if any explanation of modern attitudes to bad language is to be attempted. The sources used in this book are social and political history, sociological theory and corpus linguistics. Social and political history The British people and its government through the ages have forged the attitude to bad language current in British society today. Such a statement is clearly uncontroversial. Yet accepting this statement entails a serious examination of bad language in the context of British social and political history. This in turn leads to significant problems. Discerning the processes behind political actions and social attitudes in the twenty-first century is difficult enough. Considering such factors from the sixteenth century onwards ushers in many practical difficulties. A whole range of methodologies which may be used in the present day are clearly inapplicable when considering the sixteenth century. Focus groups, questionnaires and the full panoply of techniques in modern social science are of no use at all to the researcher in such an investigation. The limited range of data available is accessible only via the tools of the historian’s trade—dealing with old texts, government documents and whatever information other sources of documentary evidence may yield. Sociological theory It should be clear by now that my approach to bad language views it as being as much a social/historical phenomenon as a linguistic one. In trying to account for how a society develops attitudes and beliefs which problematises language, I will draw on modern sociological theory which seeks to provide an explanatory framework for such events, most notably Bourdieu’s theory of distinction and moral panic theory. Bourdieu’s theory of distinction, as will be shown shortly, is useful in explaining any differences in language use by different social classes. Moral panic theory is the basis of the approach taken in this book to discourses about bad language. Corpus linguistics Corpora are used in two distinct ways in this book. In the third part of the book, corpora are mainly used as sources of evidence to explore the development of attitudes to bad language and discourses surrounding bad language use. This contrasts somewhat with the first part of the book where corpora are used as sources of evidence related to swearing in British English. So, in the third part of this book, corpora are not being used in ways which many readers will typically be familiar with. The way corpora are used in Part 3 differs from the way in which they are used in areas more familiar with corpus use, e.g. language pedagogy, lexicography or theory-neutral linguistic description. This difference arises because my aim here is to show that corpus linguistics as a methodology allows

Swearing in English

4

one to couple corpus data with theories and supporting data from beyond linguistics. Yet in coupling corpus data with sociological theory and historical data, I believe that we gain a deeper insight into a question which should be of interest to linguists—the source and origin of the attitudes to bad language prevalent in modern British English. The first, and to some extent the second, part of the book covers a more familiar, descriptive, use of corpus data. However, it is in the contrast of the different parts of the book that I hope that the need for a deeper, historical and sociological exploration of bad language becomes apparent. While corpus data allows us to describe swearing in English, for example, it does not begin to provide an explanation for anything that we see within the corpus. Description in tandem with explanation is a powerful combination in linguistics. The separation of one from the other is damaging. An explanation of something which is not described in some credible fashion may be no explanation at all. Description without explanation is at best a first step on the road to a full investigation of some linguistic feature. In this book, corpora have a role to play in both explanation and description. The explanations for the attitudes to bad language which corpora help to flesh out in the third part of this book flow directly from the corpus-based description of bad language in the first part of the book. The explanation helps one to understand the description. The description becomes the key to lending credence to the abstract explanation. So, in this book, corpora are being used as a medium for an exploration of hypotheses arising from social and political history as well as sociological theory. Having mentioned sociological theory, it seems appropriate to return to the theories drawn on in this book: moral panic theory and Bourdieu’s theory of distinction.

Moral panics The sociologist Stanley Cohen developed moral panic theory in the late 1960s to account for episodes where the media and society at large fasten on a particular problem and generate an alarmist debate that, in turn, leads to action against the perceived problem. The response to the problem is typically disproportionate to the threat posed. Cohen (2002:1) introduces the idea of a moral panic by saying that: Societies appear to be prone, every now and then, to periods of moral panic. A condition, episode, person or group of persons emerges to become defined as a threat to societal values and interests; its nature is presented in a stylised and stereotypical fashion by the mass media; the moral barricades are manned by editors, bishops, politicians and other right-thinking people; socially accredited experts pronounce their diagnoses and solutions. Though moral panics are far from new, moral panic theory is. In spite of the relative recency of moral panic theory, it is somewhat fractured. Goode and Ben-Yahuda (1994) outline three forms of moral panic as part of an attempt to provide a grand unified theory of the topic. The problem with their approach is that it may be that in trying to produce an over-arching theory, they are forcing a separation between what may be intertwined

Bad language, bad manners

5

processes, or are forcing fundamentally different processes to sit unhappily together under the umbrella term ‘moral panic theory’. Nonetheless, as the different varieties of moral panic are of minimal relevance to the main goals and claims of this book, I will exemplify moral panic theory here solely with reference to the so-called interest group moral panic theory, both because it was the first model developed and because it links most clearly to the events discussed in Parts 2 and 3 of this book.2 Cohen (1972) put forward an early version of moral panic theory focused on a media scare related to the activities of two rival groups, ‘Mods’ and ‘Rockers’, who clashed occasionally in England, most famously in British south-coast seaside towns in 1964.3 The model put forward by Cohen is essentially a cultural account of moral panics. It has four basic elements. First, the moral panic must have an object, i.e. what is the moral panic about? Second, a moral panic needs a scapegoat, also termed a ‘folk devil’— an entity which the public can both project its fears onto and blame for a state of affairs. Scapegoats are typically vulnerable figures in the society within which the moral panic is occurring. Third, the moral panic may be generated by a moral entrepreneur via the media or by the media alone.4 Moral entrepreneurs typically represent an interest group, hence this approach to moral panics is called interest group theory. Finally, the debates prompted by moral panics are ‘obsessive, moralistic and alarmist’.5 Claims of moral decline leading to moral panics have ‘rung out down the ages’.6 In short, they are not solely a twentieth- or twenty-first-century phenomenon. One should be able to see moral panics in earlier periods of history and one should be able to fit Cohen’s model to them. Some further possible inferences that one may draw from Cohen’s work are worthy of note. First, the concept of mass media can be flexible. One need not think simply in terms of newsprint, radio and television. So, in Early Modern England the pulpit was, in effect, the mass media. In extending moral panic theory across the ages, we need to consider the changing face of the mass media over time. Second, interest group theory tends to focus on deep-seated concerns that society may hold, rather than on dayto-day concerns. In this book, when viewing a public discourse of 1699 as an example of a moral panic, I do not want to imply that if I could go back to 1699 and ask a member of the public what their main concern was that they would answer without hesitation ‘swearing in public’. Day-to-day concerns and deep-seated concerns can often diverge. It is much more likely that our interviewee would comment on some everyday need rather than on a lofty moral topic. Yet within interest group led moral panic theory, we need to explain how the interest group elevates this deep-rooted concern to a position of such importance that we might say that moral panics seem somewhat divorced from reality. In part, we can do this by saying that the interest group identifies a general concern of society and through guile or fortune manages to elevate that concern to a position of importance in the media and public consciousness. The fortune relates to the moral entrepreneur focusing on an issue which at that moment in time has become what Cohen terms a focus of cultural strain and ambiguity. The guile I include to admit the possibility that the moral entrepreneur, through the presentation of their worries, may generate a cultural strain or ambiguity. In exploring discourses of panic in Part 3 of this book, I am in part seeking to explore the guile of the moral entrepreneurs. In analysing moral panics, I claim that, within the discourse of a moral panic, there are a number of readily identifiable roles that are present across such a discourse. My development of these roles arose from a qualitative analysis of some of the moral panic

Swearing in English

6

texts in the corpora used in this book. The idea of the roles, however, arose initially as a response to my reading of the literature on moral panics. Given the features of a moral panic, as outlined in this chapter, whatever the theory of moral panic one subscribes to, there are a number of key features of a moral panic—something is identified as offensive, something or someone is blamed for this offensive thing and somebody does the accusing. In addition, the accuser often has a preferred solution to the problem, and claims that if the solution is not adopted, negative consequences will ensue. If the solution is adopted, then positive consequences will ensue. Based on these observations, I developed the following set of roles in a moral panic discourse: • object of offence—that which is identified as problematic; • scapegoat—that which is the cause of, or which propagates the cause of, offence; • moral entrepreneur—the person/group campaigning against the object of offence; • consequence—the negative results which it is claimed will follow from a failure to eliminate the object of offence; • corrective action—the actions to be taken to eliminate the object of offence; • desired outcome—the positive results which will follow from the elimination of the object of offence. In order to check the applicability of these roles to moral panic texts, I applied them to a number of texts from the corpora used in this book. The categories could be applied relatively easily to individual texts, though it should be noted that it was usually across a selection of texts from the same panic that each of the roles was filled, i.e. it is not uncommon for moral panic texts individually to represent only a subset of these roles, yet a wider set of texts from the same discourse, or indeed the discourse as a whole, will populate all of the roles in the moral panic. It is for that reason, later in the book, that large corpora containing a number of documents are used to explore moral panics related to bad language. However, to demonstrate how the roles are represented in the text, and to introduce one further category created as a result of applying the model to a range of texts, I would like to analyse one text using the model. The text in question is a letter printed in The National Viewer and Listener, autumn 1999 edition. The letter is written by a member of one of the key groups studied in this book, the National Viewers’ and Listeners’ Association, which is the focus of Chapters 5 and 7 of this book. For the moment, let me simply note that this Association campaigned7 against such things as bad language on television. The full text of the letter is given in Figure 1.1. The letter in Figure 1.1 is a single, short, example which shows the roles of the moral panic well. There is a clearly identified set of objects of offence, with sex, violence and bad language being the chief, though not the only, sources of offence identified. The object immediately responsible for the offence, the scapegoat, is television—what the children are watching, according to the letter, is harmful to them. Yet the letter also identifies a second level of responsibility—those broadcasting channels and public bodies that The power of television first impressed me when I lived near a school. Every morning as a stream of children passed by I was treated to advertising jingles, catch-phrases, unarmed combat play-acting or ‘bang, bang, you’re dead’ dialogue with bad language from the previous night’s tv programmes I began to take a

Bad language, bad manners

7

closer look at what I was watching. Did the playground echo an escalation of violence, sex and language? It led me to National VALA with its world wide findings, the concerns of others like myself and the fight to maintain common sense standards of good behaviour, decency and moral values in public communications. Ten years on the pattern has become clear. ‘Adult’ television material with its rise in violence, increasing sexual explicitness and filthy expression has abandoned responsibility for viewers of every age. Too extreme a view? Films like Natural Born Killers, Reservoir Dogs, Pulp Fiction and Trainspotting (and hundreds of similar examples shown since 1988) all on television must give any responsible citizen cause for worry. If on screen assaults, beatings, killings, shootings, woundings and brutal behaviour accompanied by revolting language and profanity and often linked with explicit sexual detail, female degradation and drugs are not considered to have a debasing influence on viewers then monitoring is pointless. But I do not think so. Knowledge has fuelled my indignation with the irresponsible response from broadcasting channels, weak regulation laid down by Government and excuses from public bodies who should know better. Good positive thinking will ensure that decency, morality and good standards return to the screen when you, the viewer, insist. After all, it is the nation and our children at risk.

Figure 1.1 A letter appearing in the autumn 1999 issue of the National Viewer and Listener. should be regulating output, as well as the Government which should be imposing stronger regulatory guidelines. These are also scapegoats. Yet encoded in the attack on the secondary scapegoats is the corrective action that the writer is seeking—the imposition of regulatory frameworks both voluntary (from broadcasting channels and public bodies) and statutory (from the Government) which would eliminate the objects of offence. This action will only occur if further corrective measures are taken, in the form of Viewers’ agitating for this change through letter writing. The claim of the letter is that, in the absence of such corrective action, there are clear consequences—the children of Britain, in particular, and the nation in general, will be harmed. Should the corrective action be taken, however, the consequences will be avoided and the desired outcome will be achieved, a Britain in which ‘decency, morality and good standards’ return to the television screen. The viewer is appealing also to an abstract moral entrepreneur—the National Viewers’ and Listeners’ Association—which is the main driver behind this particular moral panic. As a result of analysing texts such as this, I decided to introduce an additional category—moral panic rhetoric—to my analysis of the lexis of moral panics. While moral panic rhetoric is clearly different from the other categories, in that it does not identify a discourse role, it does capture an essential feature of a moral panic, as I argue that the moral panic is a distinct register marked by a strong reliance on evaluative lexis that is polar and extreme in nature. The existence of such a register is

Swearing in English

8

hinted at by Cohen (2002:19–20) when he notes, when reviewing press coverage of the ‘Mods and Rockers’ panic, that: The major type of distortion…lay in exaggerating grossly the seriousness of the events, in terms of criteria such as the number taking part, the number involved in violence and the amount and effects of any damage or violence. Such distortion took place primarily in terms of the mode and style…of most crime reporting: the sensational headlines, the melodramatic vocabulary and the deliberate heightening of those elements of the story considered as news. The regular use of phrases such as ‘riot’, ‘orgy of destruction’, ‘battle’, ‘attack’, ‘siege’, ‘beat up the town’ and ‘screaming mob’ left an image of a besieged town. While Cohen’s observations are not those of a linguist, he is clearly aware that the intentional manipulation of language to evoke specific hearer/reader responses is an intrinsic part of a moral panic, i.e. that there is a moral panic rhetoric. Indeed, in the example given in Figure 1.1, I would argue that the writer adopts moral panic rhetoric— for example, negatively loaded modifiers such as filthy, revolting, brutal, irresponsible, weak and degradation are used to amplify the objects of offence and the sins of the scapegoat. Positively-loaded words are used to describe the desired outcome that the writer and the National Viewers’ and Listeners’ Association are seeking, with talk of decency, morality, good standards, common sense and moral values establishing the moral supremacy of the writer and the National Viewers’ and Listeners’ Association and, by implication, suggesting that those who disagree with the writer are at least tacitly supporting indecency, immorality, bad standards, foolishness and the abandonment of moral values. All of these claims are based on the flimsiest of evidence—the musings of a person hearing a passing group of schoolchildren and wondering whether their behaviour might have been influenced by the previous night’s television. Rather than wondering whether the television was now more accurately portraying everyday language use, the writer chose to believe that television was setting new standards for everyday language use. Whichever of these two arguments is true, the fact that the writer does not admit the possibility that views other than their own may have validity reveals another feature of this moral panic in particular, and one that is arguably a feature of many, if not all, moral panics—the reliance on moral absolutist beliefs. As will be shown later, particularly in Chapters 5 and 7 of this book, terms such as decency and morality do not need to be defined for this writer, as they assume the meanings of these words based on a pre-existing moral framework, in the case of the National Viewers’ and Listeners’ Association, conservative Christianity. Yet the power of certainty that this gives the moral entrepreneurs and associated activists also pervades their writings—the need to explore opposing views, the need to work within a framework of moral relativism, is absent. The answers provided within a framework of moral absolutism are, by their very nature, absolute. It is that which, in part, gives strength to the rhetoric of a moral panic of this sort. Consequently in Chapters 6 and 7 I will also explore the rhetoric and discourse roles of moral panics.8

Bad language, bad manners

9

Bourdieu’s theory of distinction Another important explanatory framework adopted in this book is the theory of social distinction drawn from the work of the French sociologist, Pierre Bourdieu. Bourdieu’s work, while admittedly drawn from his research on French society and relating largely to features of culture such as art, food and manners, nonetheless is relevant to language, as Bourdieu himself acknowledges. Bourdieu’s claim is a relatively simple one: features of culture are used to discriminate between groups in society, establishing a social hierarchy based on a series of social shibboleths. The consequences of the establishment of such a hierarchy are both to allow members of groups to be readily identified and to impose the hierarchy itself. For example, if a taste for fine wine is supposed to be a token of high social status, then on seeing somebody pouring a drink from such a bottle of wine, other factors aside, one might assume they were of a certain social class. Similarly, if one sees somebody drinking a pint of beer, and this is a marker of low social class, other factors aside, one may also infer their social class. However, if fine wine is priced so as to exclude the lower orders from purchasing it, the social hierarchy has nothing to do with taste as such. Rather, those tokens of taste are controlled in such a way as to impose the social structure that they are a token of. Transporting this argument to language is somewhat straightforward. If there are forms of language which are identified with a refined form of speech, then those aware of the perception of this form of language, who are able to invest either the time or the money in order to acquire that ‘refined’ form of language, will be able to identify themselves with a particular group in society. Yet more perniciously, if that type of speech is already associated with a particular social class, then there is a zero cost for that social class in using that form of speech, while the speech associated with lower classes is devalued and the onus is placed on them to adapt the way that they speak. In making that adaptation they are tacitly acknowledging the supposedly superior form of speech that they are shifting to when that shift takes place. To Bourdieu, in language this process leads to: opposition between popular outspokenness and the highly censored language of the bourgeois, between the expressionist pursuit of the picturesque or the rhetorical effect and the choice of restraint.9 In seeking shibboleths of taste, groups distinguish themselves from one another in society in order to set boundaries which identify difference. For Bourdieu this means that: Groups invest themselves totally, with everything that opposes them to other groups, in the common words which express their social identity, i.e. their difference.10 In other words, the process of setting out the boundaries of linguistic differences for groups is no casual process. It is a process whereby the very identity of the groups concerned becomes intimately associated with their language use, through ‘the socially charged nature of legitimate language’.11 Linked to a social hierarchy, the capacity is clearly generated to identify not merely the language of particular groups, but to identify the language of various groups with power as defining a discourse of legitimacy, a

Swearing in English

10

discourse of power. This discourse of power then becomes the unmarked case—the linguistic norm, the supposedly neutral form of expression—with forms that do not follow it marked out as the marked, abnormal, negatively charged forms of language, or ‘the least classifying, least marked, most common, least distinctive, least distinguishing’12 forms of language. This process of the discourse associated with one group becoming the dominant discourse of power leads to those not possessing that discourse being: at the mercy of the discourses that are presented to them… At best they are at the mercy of their own spokesmen, whose role is to provide them with the means of repossessing their own experience. The essential indeterminacy of the relationship between experience and expression is compounded by the effect of legitimacy imposition and censorship exerted by the dominant use of language, tacitly recognized, even by the spokesmen of the dominated, as the legitimate mode of expression of political opinion. The dominant language discredits and destroys the spontaneous political discourse of the dominated. It leaves them only silence or a borrowed language.13 In other words, those without access to this discourse of power are already marked as disadvantaged by their language use. This disadvantage is compounded by them having to use a discourse with which they do not readily identify when asserting themselves, as: Through the language… Bound up with a whole life-style, which foist themselves on anyone who seeks to participate in ‘political life’, a whole relation to the world is imposed.14 At worst it may lead to the failure of the dominated groups to represent themselves, relying rather on members of the group possessing the dominant discourse consenting to represent them and provide leadership to them, as Bourdieu notes when he says that: It forces recourse to spokesmen, who are themselves condemned to use the dominant language…or at least a routine, routinizing language which…constitutes the only system of defence for those who can neither play the game nor ‘spoil’ it, a language which never engages with reality but churns out its canonical formulae.15 Distinction simultaneously empowers further those already possessing power, while further dispossessing those who are already dispossessed. This book will argue that, when we look at modern English, we see distinction at work in the form of bad language. Broadly speaking, the discourse of power excludes bad language, the discourse of the disempowered includes it. Obviously, this statement is, however, something of an idealisation, as several factors may, for example, combine on any specific occasion to determine language usage. Similarly, several factors together may establish a matrix of power, as opposed to single factors generating a polar distinction between the powerful and the disempowered. Indeed, in Part 2 of this book I will explore how demographic factors may combine in such a way. For the moment, I will maintain the broad assertion

Bad language, bad manners

11

made above, adding the caveat that such a statement notes what is typical and is only generally applicable when we are considering one feature in isolation. One final point I should make at this stage is that what I am discussing here is overt as opposed to covert prestige.16 In this book I am mainly concerned with power related to overt prestige, though I accept without hesitation that in establishing an overtly prestigious form of language, a covertly prestigious form of language is entailed which may invert the matrix of power mentioned above. Research into overt and covert prestige is so well established that I feel the issue can be sidestepped in this book as there is a wealth of material that interested readers can pursue to explore this issue for themselves.17 To recapitulate the earlier goals and claims of this book, it will be argued that the process of forming a class distinction around swear words was undertaken in the late seventeenth/early eighteenth century by an aspiring middle class who actively sought to distinguish themselves from the lower orders by a process of ‘purifying’ the speech of the middle class while prob-lematising the speech of the lower orders (see Chapters 4 and 6). Further, it will be argued that the vehicle which brought about this process of distinction was a moral panic focused on bad language in the late seventeenth century, which empowered certain members of the middle classes to act simultaneously as moral entrepreneur and arbiter elegantium, dictating the linguistic manners of the general population. Finally, the book will argue that the processes of disempowerment which Bourdieu suggests are entailed by such a development are observable not merely in the seventeenth century but in the present day (see Chapters 5 and 7). This brief overview of the book allows readers to see how the elements introduced in this chapter come together in order to provide a coherent account of bad language in English. The corpus is used principally to establish a series of observations related to distinctions in the use of swearing. The explanation of these distinctions is then sought through historical research, as well as the application of moral panic theory and Bourdieu’s theory of distinction to texts in the period 1690–1745. The process of disempowerment is then explored further in the context of debates about language in the media in the late twentieth century. While corpus data will be instrumental in exploring the discourse of bad language in the seventeenth, eighteenth and twentieth centuries, it is the goal of this book to show that explanations for what we see in corpora often lie beyond the borders of the corpus itself—the observations we can draw from corpora, while verifiable, are not necessarily of any assistance in developing explanations, though they do frame what an acceptable explanation may look like, i.e. any explanation must match the observations drawn from the corpus. But by marrying other methodologies with the corpus method, and drawing on appropriate theories, the corpus data itself can be illuminating in the search for a wider, comprehensive account of the features of language we approach the corpus to investigate.

Corpus linguistics: the corpora used in this book In this section, I will discuss the majority of the corpora used in this book. Two minor corpora (a corpus of seventeenth-century news texts and a corpus of German radio propaganda broadcasts) will be discussed briefly when they are introduced. One major

Swearing in English

12

corpus, the Lancaster Corpus of Abuse,18 is not reviewed in this section, being reviewed instead in Chapter 2 as a prelude to an analysis of bad language in present-day English. The Mary Whitehouse corpus (MWC) The MWC includes the major writings of Mary Whitehouse in the period 1967–1977. This corpus covers three of her books, namely Cleaning-up TV, Who Does She Think She Is? and Whatever Happened to Sex?, amounting to 216,289 words in total.19 These books, with their wide circulation, were the principal public output from the National Viewers’ and Listeners’ Association (the VALA—see Chapter 5 for details) in this period, and as such I take them to be a good focus for a study of how the VALA tried to excite a moral panic in the general population of Britain. The British National Corpus (BNC) The BNC is a 100,000,000-word corpus of present-day British English. The corpus is split into a 90,000,000-word balanced written corpus and a 10,000,000-word corpus of orthographically transcribed spoken language. As I am using only the spoken data in this book I will limit my brief description of the BNC to its spoken section. The spoken BNC is composed of a series of spontaneous conversations recorded by members of the British public in the early 1990s. The corpus was designed to provide material from across the UK (the so-called demographically sampled subset of the corpus) and across a range of different activities (the so-called context governed subset). Demographic information about the speakers was encoded in the corpus. This demographic data was then used to balance the spoken material with regard to a number of variables, notably, for this book, age, sex and social class. The result of this balancing is that, in the corpus, the amount of speech spoken by males and females is roughly even, as is the speech produced by different age groups and social classes. The Society for the Reformation of Manners corpus (SRMC) The SRMC was compiled by me specifically for this study. It contains four key texts from the Society for the Reformation of Manners (SRM) amounting to 120,709 words.20 Two texts were selected as being those which achieved the widest circulation during the period of the reformation of manners movement and which were widely cited—Yates (1699) and Walker (1711)—while two further texts were included from the end of the period of the society’s activities, namely Anon. (1740) and Penn (1745).21 The latter texts were included to permit an investigation of how, if at all, the discourse of the society shifted during its lifetime. While ideally one would like to have gathered a much larger set of texts together, the longevity of the Yates and Walker texts, and their wide distribution during the lifetime of the societies, makes them in essence texts which are representative of the society and its aims. The later texts, as noted, represent some of the final texts of the society and are included solely to allow the possibility of a diachronic approach to the writings of the society.

Bad language, bad manners

13

The Lampeter corpus The Lampeter corpus is a diachronic corpus of English, covering the period 1640–1740. The corpus samples texts from a range of genres (economy, law, miscellaneous, politics, religion and science) over this period, taking samples at periods of roughly ten years. The corpus was constructed at the University of Chemnitz by a team led by Josef Schmied, and has been used in the diachronic study of variation in English.22 For the purposes of this book, I will only use materials from the corpus covering the period 1690–1750, as it is in this period that I want to contrast the language of the SRM with what one might term English in general23 (i.e. all of the genres of the Lampeter at once) and specific genres and registers of English (texts covering only one domain of Lampeter). There are 544,894 words in the Lampeter corpus in the period 1690–1750. The Lancaster—Oslo—Bergen and Freiberg—Lancaster—Oslo— Bergen corpora (LOB and FLOB) Both the LOB and FLOB corpora are related to an earlier corpus, the Brown University Standard Corpus of Present-day American English (i.e. the Brown corpus, see Kucěra and Francis 1967). The corpus was compiled using 500 chunks of approximately 2,000 words of written texts. These texts were sampled from 15 categories. All were produced in 1961. The components of the Brown corpus are given in Table 1.1. LOB and FLOB follow the Brown model. The Lancaster—Oslo—Bergen corpus of British English (LOB) is a British match for the Brown corpus.24 The corpus was created using exactly the same sampling frame, with the exception that LOB aims to represent written British English used in 1961. The Freiberg-LOB corpus of British English (i.e. FLOB) represents written

Table 1.1 Text categories in the Brown corpus Code Text category

No. of samples

Proportion (%)

A

Press reportage

44

8.8

B

Press editorials

27

5.4

C

Press reviews

17

3.4

D

Religion

17

3.4

E

Skills, trades and hobbies

38

7.6

F

Popular lore

44

8.8

G

Biographies and essays

77

15.4

H

Miscellaneous (reports, official documents)

30

6.0

J

Science (academic prose)

80

16.0

K

General fiction

29

5.8

L

Mystery and detective fiction

24

4.8

Swearing in English

M

Science fiction

N

14

6

1.2

Western and adventure fiction

29

5.8

P

Romantic fiction

29

5.8

R

Humour

9

1.8

500

100.0

Total

British English as used in 1991 using the Brown sampling frame once more.25 LOB and FLOB, as well as being corpora which allow one to study recent change in British English, may also be used, as they are in this book, to exemplify general published written British English in the early 1960s and early 1990s respectively. Issues Before leaving the presentation of the corpora used in the book, it is appropriate to pause and consider a number of methodological questions arising from the use of the corpora. The first relates to claims of balance and representativeness for the specialised corpora, i.e. the MWC and the SRMC. In what way might they claim balance and representativeness? Clearly not in the same way as the BNC or Lampeter corpus can. For example, the SRMC is not particularly representative of general English in the period in which it was written. Similarly, it is not balanced with regard to general English in the period. But both of these assertions of course miss the point—it is not intended to be generally balanced and representative; it is not representing general English in the late seventeenth century. Rather, it is representing the writings of a specific group in that period. So balance and representativeness for the MWC and the SRMC should relate only to the writings of the group or writer in question, not for writers of the language in general. Yet, the focus of the specialist corpora is narrower still. The corpora in question are not trying to be representative of all of the works produced by the group in question. If that were the case, such items as handbills handed out by the Society for the Reformation of Manners to those it had had prosecuted would have to be represented.26 But the purpose of the study of the SRMC in this book is to explore the way in which the SRM attempted to persuade society at large of its case, and more specifically how they sought to persuade society that bad language was a major problem. As such, the SRMC was constructed to focus principally on those texts which achieved very wide circulation in Britain, as one may hypothesise that it was these texts, rather than handbills handed to individuals, that had the greater impact on British society. A similar argument applies to the MWC—it is by studying the widely published works of Mary Whitehouse that we can see the effect on discourse that the VALA had, not by looking at newsletters produced for the relatively small number of subscribing members of the VALA. Yet, for both the SRM and the VALA, I would not want to claim that the more ephemeral texts they produced had no impact on society—I am sure that their effect on a micro level was notable. However, as throughout the rest of this book I am trying to focus on macro rather than micro processes with regard to attitudes to bad language, and hence the major, widely disseminated texts of the SRM and the VALA are the focus of the corpora built for this book. Balance and representativeness as issues become somewhat narrow when one has

Bad language, bad manners

15

such a tightly focused research question as that under consideration here. A more fruitful way of approaching the specialised corpora constructed for this book is to think of them as corpora which are focused on a very narrow issue. For such corpora, the concepts of balance and representativeness become so specific as to be uninteresting. A slightly more difficult issue relates to the question of the comparability of the reference and the specialised corpora, for example Lampeter with the SRMC.27 The timeframe for the Lampeter corpus is 1640–1740. For the SRMC it is 1699 to 1745. Can one truly compare the SRMC and Lampeter with confidence when the Lampeter corpus cannot match the timeframe of the SRMC texts perfectly? There are two responses to this problem, one pragmatic, one principled. The pragmatic response is that, at times, the perfect corpora for any given study may not exist, but one may still proceed to undertake an exploration with imperfect corpora as long as one notes that, at some future point when the perfect corpora are available, researchers may wish to return to the results in order to verify, in this case, that the differences seen between the corpora were a result of a process other than language change. I encourage future researchers to do just that, as I am working with the best corpora available to me, and accept the possibility that future corpora may reveal that the differences noted in this book have everything to do with language change and nothing to do with moral panics. However, I doubt that this will happen, because of the principled point I want to make. While some features of language change rapidly—notably lexis—other features of language change much more slowly, for example, grammar. Relatively large corpora covering relatively large time periods are needed to catch grammatical change. I liken the process to normal cinematography as opposed to time-lapse cinematography. A corpus like the BNC is like a typical camera— it may be useful for capturing movements which are relatively rapid. A pair of corpora like the LOB and FLOB corpora, which are two corpora with identical sampling frames applied to English in the early 1960s and early 1990s respectively, are needed to catch much slower movements not immediately visible using a corpus such as the BNC.28 Just as we need to use time-lapse photography to see a flower open, so we may use carefully sampled corpora with identical balance and representativeness, built to represent the same language, across a significant period of time to see slow-moving language change. My view is that the changes being looked at in this study are relatively slow moving— discourses of moral panic, I will claim, have fairly stable properties, so much so that over nearly 300 years we can see marked similarities between the panic discourse of the SRM and the VALA. Given that we are dealing with stability over time on such a scale, I think one can fairly view the slight differences in timeframe between the focused and reference corpora used in this book as being largely irrelevant. One final issue I need to deal with is the question of variant spelling in the Early Modern period. For example, variant spelling occurs in the SRMC in the sense that certain words may be spelt differently in different texts, or even at times within the same text. Also, though at times a word form has a relatively stable spelling in the texts, the word form used is not identical to the modern English word form (e.g. publick v. public). The former type of spelling variation in particular can cause problems when exploring word frequency, as the word may be represented by many word forms, each with a separate frequency. As word frequency is an important measure used in this book, the issue of spelling variation had to be addressed. Consequently, when constructing the SRMC corpus, where spelling variation occurred, both the original word form (e.g.

Swearing in English

16

govenour) and the modern word form (e.g. governor) were encoded in the text. While this may not be of interest to the average reader, I should note that by the use of a markup language called XML I was able to encode this information in the corpus in such a way as to allow readers/analysts to see either the original spellings in the text or the modern variants, as they wished. Throughout this book, I will use the modern spelling forms encoded in the texts to construct the wordlists used in my study.

Concepts/techniques used in this book Word frequency forms the backbone of the analyses undertaken in this book. The reason for my focus on word frequency is related to an observation made on page 6—the discourses surrounding moral panics are obsessive, moralistic and alarmist. Each of these factors should, in principle, significantly influence the lexis of the moral panic. Moralistic and alarmist language should be signified by words with an alarmist or moralistic tone such as, perhaps, evil, threat or danger. My hypothesis is that these words will be used more frequently by purveyors of moral panic, and that the obsessive nature of their discourse should lead to the use of these words becoming not merely frequent, but so frequent that these words can be viewed as salient, in the sense that they distinguish texts relaying moral panics from general English, or even texts written in a similar register/genre but not conveying a moral panic. Similarly, the obsessive focus of the text on particular problems, solutions and scapegoats should mean that words denoting these elements of the moral panic will also become salient in the text. In order to explore moral panics in this way, I used the keywords function of a computer program called WordSmith29 in order to find those words which occurred in the focused corpora significantly more frequently than in the reference corpora. As the keywords analysis is so central to the study presented in this book a brief discussion of the workings of keywords is necessary.30 Keywords and key-keywords Keywords, as conceived by linguists,31 are those words which, when a particular corpus (A) is compared to a reference corpus (B), are used significantly more or less frequently in A than in B. Note that the choice of A and B to a large extent determines how we can interpret the results of a keyword analysis. Imagine that A is a highly specialised type of language, for example computer manuals, while B is a collection of written language similar to the written section of the BNC, i.e. a balanced and representative sample claiming to represent general English. Such a comparison will most likely show the differences between the specialised variety of English and English in general. We would expect words like monitor, mouse and keyboard to occur much more frequently in computer manuals than they generally would in written English. These are termed positive keywords, as they represent some of the lexis which is used more frequently by writers of this type of text, and hence may be said in a way to characterise this type of text. Similarly, words which may occur in written English quite frequently, such as car, laugh or stroll are clearly less likely to occur in computer manuals and hence may show up as negative keywords—lexis that is shunned by the writers of computer manuals.

Bad language, bad manners

17

Positive and negative keywords may also tell us something about the nature of the discourse in the texts other than the topic area under discussion—pronouns, for example, may appear as keywords. The computer manual example will serve to illustrate this point again. It is likely that computer manuals, written to provide instructions to users, will include a higher than usual proportion of second-person singular pronouns, as you will be used to direct instructions to the reader. On the other hand, the other pronouns, which may be fairly evenly spread across the reference corpus, will show as negative keywords in the corpus as the computer manual texts will shun them. Keywords are determined by WordSmith using a test for significance called the pvalue,32 which is calculated in this book on the basis of the log-likelihood score.33 Words become keywords if they are used with a difference in frequency between corpora A and B in such a way that their frequency in A is significantly higher than B (positive keywords) or lower than B (negative keywords). In addition, the keywords themselves are ranked by the WordSmith program and given a keyness score to denote the strongest to weakest negative and positive keywords. While comparing a specialised form of text to a general corpus of English is bound to achieve some fairly obvious results, this is not the only valid use of keyword analyses. We may, for example, wish to compare apparently similar types of texts to one another in order to identify relatively subtle differences between those texts. Consider a situation where you have access to the writings of two newspapers and you wish to see if there are any differences between the two. Setting aside the possibility of comparing radically different newspapers, let us consider what might happen if we compare two quality broadsheet newspapers, for example the New York Times and the Los Angeles Times, or The Times of London and the Independent newspaper from the UK. One would reasonably hypothesise that a number of differences could be shown between the papers concerned. It is not likely that we would see the genre specific lexis appearing as positive keywords, as happened in the fictitious comparison of the BNC and some computer manuals. Similarly, it is unlikely that the negative keywords would contain much in the way of general English in use by one newspaper but not the other. Rather, we might see lexis which betokens differences in editorial style perhaps (one newspaper may have house style rules that dictate that first person pronouns are not to be used, the other may have no such rule leading to I becoming a keyword) or perhaps related to differences in reporting practices (one newspaper covering a particular issue regularly in depth more than the other newspaper), or perhaps differences brought about because of the place of publication (perhaps California may be a key word in the comparison of the LA Times and the New York Times simply because one newspaper was published in California and reports the news from California in more depth as a consequence). So comparing texts which appear similar may be as rewarding as comparing texts which are obviously different, though the results of the analysis will most likely be somewhat different.34 The reason that I spent some time discussing negative and positive keywords, and their use in studying similar and different text types, is that in Part 3 of this book keyword analyses of this sort will be undertaken, and the positive and negative keywords generated by these analyses will be the main focus of the exploration of the moral panics encoded in the SRMC and the MWC. I hypothesise that the keyword list, when one compares one of the specialised corpora to its corresponding reference corpus, should be populated, in part at least, with words that identify the major roles and actors of the moral

Swearing in English

18

panic, as these should occur with an unusually high frequency owing to the obsessive, alarmist and moralistic nature of the moral panic. A refinement of the keyword analysis, key-keywords, is also used in this book. Keykeywords are keywords which are key in all, or the majority, of subsections of a corpus. I use the term subsection here as the calculation of keywords is actually undertaken across individual corpus files. These may represent almost anything, e.g. a vast collection of different texts held within one file, an individual text, or a fragment of text.35 However, assuming that the files in a corpus represent texts in some meaningful way, key-keywords are of particular use in exploring such issues as whether a keyword is key to a colony of texts (where each file is a text) or across a whole text (where each file, for example, is a chapter of a text). Key-keywords are used in Chapter 7 of this book. Collocates, linked collocates and colligates There is one other major form of analysis which this book will use which should be introduced here: collocational networks. In part, I will be exploring collocational networks to look for linked keywords, keywords which are linked by common collocates, or, as I will term them, link collocates. Generally, I will use collocational networks to pursue the lexical organisation of text along the lines suggested by Martin Phillips.36 Collocations are explored in this chapter, using the mutual information statistic as a useful heuristic to filter meaningful from non-meaningful collocates. Collocation is the process whereby words keep company with one another and thereby convey meaning via co-occurrence. The idea is not particularly new,37 though collocates are still being analysed, refined and explored by corpus linguists.38 For ease of discussion, some basic terminology used with reference to collocation needs to be introduced beginning with node and span. A collocate, for the purposes of this discussion, is a word which occurs with some higher than chance frequency in the context of a given word. The given word is called a node and the span of words either side of the node in which we search for collocates is termed the span. In this book, a span of five words either side of the node will be used in exploring collocation.39 A distinction between collocation, a frequent association with content words, and colligation, frequent association with grammatical words, is sometimes drawn by researchers. While this distinction is noted and accepted by this work, the distinction is not particularly active in the analysis of the corpora used in this book, though where it is, the distinction will be sustained. Collocational networks Given a working definition of collocation, we need to consider a number of known properties of collocates. First, collocations are directional.40 For example, while we might observe red collocating with herring, the association of herring to red is much stronger than the association of red to herring. In short the link between the two may be seen as more important to herring, which, when it occurs, is most likely to co-occur with red, than to red, which has a wide number of collocational partners, of which herring is one. Second, certain words attract more collocates than others. While the specific words attracting collocates may vary across a range of written contexts, given the means to investigate specific texts and corpora, we will find words which establish networks of

Bad language, bad manners

19

collocation. Within these networks, the words which attract most collocates to them, or are in some sense central to that collocational network, are called nuclear nodes. Figure 1.2 is an example of the word POINT41 occurring as a nuclear node in a thermodynamics textbook.42 In this work I will at times be using collocational networks to explore patterns of lexis surrounding certain keywords and other specific node words central to the arguments put forward in this book, e.g. SWEAR, LANGUAGE. I will focus only briefly, and rather technically, here on the extraction of such networks, as I wish to keep the focus on the uses that such networks may be put to, and how my use of them differs from that of others. Readers uninterested in the precise technique for extracting these clusters are advised to skip to page 24 at this point.43 The construction of collocational networks undertaken in this book

Figure 1.2 A sample collocational network. groups words together on the basis of the strength of the association of collocates with a given node. In order to determine if the link between a candidate collocate and a node is strong enough for the two to be linked, mutual information is used.44 Mutual information (MI) measures how often, in a given corpus, words are attracted to one another relative to their occurrence independent of one another. In relative terms, if the measure produces a positive score, the words are attracted to one another (they co-occur frequently), if the score is around zero, the two words in question have no particularly strong association, while if they yield a negative score they shun one another’s company. In this study, I will include a link between two node words where the MI score exceeds 3.45 I will indicate with an arrowhead the direction of association, where appropriate. Where no arrowheads are shown, the link can be assumed to be bidirectional. By way of illustration, Figure 1.3 shows the pattern of collocation focused around the word swearers in the SRMC, as explored in Chapter 6. It should be noted that MI is not a rigorous statistic. It is certainly not a parametric test. One cannot reasonably talk about ‘statistically significant’ results being produced by MI. However, MI is a very useful heuristic which describes data and helps in the process of interpreting complex data sets like large corpora. It is in this spirit that MI, and the collocational networks based on this measure, are used in this book. Nonsense collocational networks, clusters and MI scores, in the sense that they defy reasonable

Swearing in English

20

explanation, can and do occur. However, the MI measure is helpful many times more often than it is unhelpful, and as such it provides a powerful tool to the linguist interested in studying patterns of collocation. The interactions observed, of course, require interpretation, as will be shown later, because the networks may describe a range of behaviours—such as words participating in the creation of terms or the different meanings of a word (the example given in Figure 1.2 was used by Phillips to demonstrate this property of collocational networks with reference to the word point).

Figure 1.3 The network around swearers in the SRMC. Semantic prosody A further thing that I argue a collocational network may help to show is the semantic prosody of a particular word. As Stubbs (2002:225) notes, ‘there are always semantic relations between node and collocates, and among the collocates themselves.’ The meaning arising from the interaction between a given node word and its collocates is referred to as semantic prosody, ‘a form of meaning which is established through the proximity of a consistent series of collocates’ (Louw 2000:57). Semantic prosodies typically convey meanings that encode attitudes and evaluations (Louw 2000:58). Semantic prosodies are typically negative, with relatively few of them referring to an affectively positive meaning. Semantic prosody is strongly collocational in that it transmits meaning beyond the sense of individual words, i.e. words which do not convey a negative meaning in isolation convey one when they collocate together. I will not discuss semantic prosody in more depth here, as it is used, and to an extent exemplified, in Chapters 6 and 7. The use of the techniques My aim in using the techniques described here is to access both the aboutness of the individual texts and collections of texts used in Part 3 of this book. I also wish to show

Bad language, bad manners

21

the patterns of meaning being formed within a text related to certain concepts in the text. To deal with the first point, I want to be able to characterise the texts generally, in essence asking the question, ‘What is this text about?’ through largely automated means, as a prelude to exploring the moral panics encoded within the texts. As part of exploring the moral panics I believe to be encoded within the texts, I also wish to pursue the question of the collocational networks surrounding particular node words, accessing these networks in order to demonstrate how the meaning associated with that node is constructed. What company does a word such as swearing keep? How is its meaning coloured by its association with that company? Are collocational networks a means of exploring how attitudes to swearing have been formed and reinforced? It is in response to questions such as these that I will be exploring specific collocational networks (Chapter 6). I will also work with the corpus on a number of levels; I will explore the corpora at the level of the whole corpus, whole texts within a corpus and whole chapters within a corpus (see Chapter 7).46 The overall characterisation of the aboutness of the corpora will be undertaken using a keyword analysis and exploring the links between the keywords. The exploration of the keywords will be supported by looking at the collocational networks focused around certain keywords. Each keyword, for the purposes of this book, will act as a node.

The book in outline With an outline of the main theories to be used in this book provided, and my methodology established, we can now consider the way in which theory and methodology will come together in this book. The book is divided into three parts. Part 1 consists of Chapter 2, Part 2 consists of Chapters 3 to 5 and Part 3 consists of Chapters 6 and 7. Part 1 is an investigation of how bad language is used in present-day English. The focus for the study is the spoken section of the British National Corpus. Using the corpus, the relationship between bad language and a number of social variables, namely age, sex and social class, is examined. Part 1 concludes by relating the differences found in the corpus to Bourdieu’s theory of distinction. Having observed a number of social variables interacting with bad language in Part 1, Part 2 will set out to explore the historical context in which the distinctions apparent in the spoken BNC developed. In doing so, the chapter will explore a period before which these attitudes developed (Chapter 3), showing how bad language was not subject to widespread, state sponsored regulation, nor was it particularly associated with age, sex or social class. Chapter 4 explores the social processes whereby distinction became focused on bad language, generating the links with age, sex and social class observed in Chapter 2. In doing so, the chapter will begin to link moral panic theory and distinction, claiming that a moral panic about immorality led to a wider social movement which caused bad language to become a marker of distinction. Chapter 5 follows on from Chapter 4, moving the discussion of bad language and morality to the late twentieth century by exploring reactions to the use of bad language in the popular media in the 1960s and 1970s. In doing so, the chapter refines the theory of distinction further by relating it to a model of discourse in which purity has become equated with power, allowing for the possibility that power may be undermined by

Swearing in English

22

deliberate verbal impurity. In exploring this issue, the chapter focuses very much on the campaigns of Mary Whitehouse against bad language in the UK through her National Viewers’ and Listeners’ Association. Part 3 reflects back on Part 2 by exploring the discourses of moral panic evident in the writings of the groups covered in Chapter 4 (explored in Chapter 6) and Chapter 5 (explored in Chapter 7). In exploring the discourses of these panics, the chapters investigate both specific rhetorical devices used to produce panic, form attitudes and assert the existence of in and out-groups in the societies being discussed. The three sections together explore bad language use now (Part 1) the historical roots of current linguistic usage (Part 2) and discourses that have influenced that usage over time (Part 3). In terms of the use of corpora in the book, their role is crucial in Parts 1 and 3. In Part 1, it is corpus evidence which outlines the patterns of bad language use, which the historical account of the development of attitudes to bad language in Part 2 must explain. In turn, in Part 3 the discourses which are explored using data are explored in the light of the historical and social arguments presented in Part 2. Throughout, sociological theory is used to account for what I would argue is an essentially social process—the association of certain words with certain variables such as age, sex and social class. It is in using corpus data to control and direct historical, linguistic and sociological enquiry that I hope this book can prove to be thought-provoking.

Part 1 How Brits swear47

2 ‘So you recorded swearing’ Bad language in present-day English48 Bad language as a marker of distinction Bad language words (henceforth BLWs) are a marker of distinction in English. As will be shown, distinction, BLWs and a range of sociolinguistic variables interact in ways which are at times predictable. Yet, at other times, they are quite unexpected. This chapter will explore those patterns of interaction. Before doing so, however, I feel I should address my assertion that BLWs are a marker of distinction as opposed to simply being markers of difference. I accept that one might normally express observations such as those that will be made in this chapter in terms of difference. Yet in the case of BLWs I argue that what we are in fact looking at is a process of distinction—the difference is directly related to prestige. BLW use is one of a number of linguistic variables we may consider when we discuss non-prestige forms of language. Hence the presence or absence of BLW use in language is a marker of distinction, with the relative absence of BLW use being a marker of a more prestigious, more refined version of the language. I will leave this defence of my approach to BLW use being based on distinction here for the moment— the conclusion to this chapter, Chapter 4 and Chapter 5 return to this issue. The conclusion of this chapter refines the notion of distinction in the light of the findings here. Chapters 4 and 5 review two discourses about BLW which very clearly make the point that BLW is linked to prestige. In this chapter I will look at BLWs in English, as used in everyday speech, in order to explore the ways in which distinction relates to it. In doing so, I will explore the behaviour of single BLWs, groups of BLWs and types of BLWs. We will see how those words are related to specific groups, or may be indicative of interactions between specific groups. However, before exploring these issues, let me present the data used in this chapter.

The Lancaster Corpus of Abuse The work in this chapter is based on the Lancaster Corpus of Abuse (LCA) which in turn is based on the BNC spoken corpus. The LCA is a problem oriented corpus based on data extracted from the BNC spoken corpus.49 The corpus contains only those examples of BLW usage where the age, sex and social class of the speaker are known.50 Within the corpus, BLWs have been annotated using a scheme developed to encode a range of information relevant to the linguistic study of such terms. In this chapter, when I am

‘So you recorded swearing’

25

looking at these features—notably categories of BLW use and gender of the target of a BLW—I use the LCA. Otherwise I use the whole BNC spoken section in order to increase the volume of examples retrieved. In deciding what words I wanted to include within the corpus, I was partly guided by claims within the literature, partly by my own intuition, partly by serendipitous discovery and partly by words I encountered within the corpus which fitted the classification system developed. This later point will be returned to shortly. To give examples of each of the first three types of words, I have used the corpus to explore claims made by researchers such as Hughes (1998) in his work on swearing in English.51 Hence, words explored by Hughes were included in the corpus. Yet I knew, on the basis of my own knowledge of bad language, that there were many examples of bad language not addressed in the literature. I then expanded the coverage of the LCA (2.0) on the basis of my intuition. Beyond this, however, when I examined the corpus data from time to time I would come across new examples of bad language. Sometimes these words were familiar to me, but I had simply forgotten them (e.g. pissy). However, sometimes the words/phrases were entirely new to me and their discovery by me in the corpus was entirely accidental (e.g. battyman as an abusive term to refer to male homosexuals) and sometimes they were relatively novel terms formed, for example, as a result of word play (e.g. Cuntona occurs in the BNC as an insulting pun on Cantona, the surname of a French footballer). The LCA inherits the balance of its parent corpus, the BNC, which is balanced for age, sex and social class. This is fortunate, as these are the three variables with which this chapter is principally concerned. The BLWs covered by the LCA can broadly be grouped under the following main headings—swear words (e.g. FUCK, PISS, SHIT), animal terms of abuse (e.g. PIG, COW, BITCH), sexist terms of abuse (e.g. BITCH, WHORE, SLUT), intellect-based terms of abuse (e.g. IDIOT, PRAT, IMBECILE), racist terms of abuse (e.g. PAKI, NIGGER, CHINK) and homophobic terms of abuse (e.g. QUEER). Obviously, there is an interplay between these broad categories—for example, animal terms of abuse may also be sexist abuse forms (e.g. cow). However, for the purposes of describing the contents of the corpus, this broad classification will suffice.52 All told, there are 8,284 separate examples in the LCA. The corpus also contains annotations, as in order to retrieve information from the LCA in a speedy and systematic way,’ I needed to develop a set of annotations geared to the study of BLWs. In the following section the development of this system will be outlined. The annotation scheme The examples in the LCA are all annotated so that the relevant metadata encoded in the BNC is retained by the example in the LCA. So, for example, if an utterance in the BNC was spoken by a male, aged 0–15, of social class DE, this information is retained by the LCA. Note that in building the LCA I was only interested in examples for which all of the relevant metadata was available. For example, where the age, sex or social class of a speaker of a BLW was unknown in the BNC, this data was not included in the LCA, as my purpose in building the LCA was to develop as rich a set of annotated data related to BLWs as was possible. Note, however, that in the study of BLWs related to the age, sex and social class of speakers in this chapter, the whole BNC is once again used to

Swearing in English

26

maximise the number of examples recovered, so the restrictions imposed on the LCA are only relevant to a subset of the features explored in this chapter. One major limitation imposed on the LCA was inherited from the BNC. As well as studying who spoke an instance of BLW use, I was equally interested in knowing who they had been speaking to. However, with the spoken BNC it is impossible most of the time to discover who the hearer/hearers of any given utterance was/were. Similarly, it can be very difficult on occasion to determine the gender, age and social class of the person or object a specific example of BLW use is directed at. Hence while the LCA represents a richly annotated corpus for the study of BLWs, the corpus does not represent the ideal resource with which to investigate BLWs. It does, however, represent the best alternative given the resources available to construct it. While the LCA owes a great deal of its annotation to the BNC, as it simply inherited annotations from that corpus, additional annotations were introduced to the LCA data, namely gender of target, animacy, metalinguistic usage and BLW type. I will describe each of these briefly. Where it was possible to determine the gender of the person at whom any example of abusive language has been directed, that was annotated. The identification of such examples relied on a close reading of the context of the utterance, with clues, particularly gender marked pronouns, being very important in the process of determining the gender of the person at whom the bad language was directed. Similarly, information on the animacy of the object/person at whom the bad language was directed was annotated in the corpus—again this required a human analyst to read the text to determine the object/person in question. Where it was possible to identify at who/what the example of bad language was directed, the animacy of that object/person was then noted. Metalinguistic usage was also noted—not all examples of bad language in the BNC constitute the actual use of bad language. On occasion people are merely discussing their favourite swear word, for example. By considering each example in the corpus, such metalinguistic uses of bad language were identified. Finally, a scheme which categorised the type of bad language in use in each example was developed. I will not discuss this classification scheme in detail here as it has been the subject of discussion in a number of other publications.53 However, suffice it to say that the scheme itself has undergone a number of changes since its inception and that some categories, especially those related to metaphoric usage, are certainly susceptible to further development. With that said, the scheme itself proved robust when it was manually applied to the corpus, and as such seems to provide a credible basis for the categorisation and differentiation of the uses of BLWs. Table 2.1 outlines the bad language categorisation scheme, while Table 2.2 shows the full range of annotations applied to each example of bad language in the LCA. A brief discussion of Table 2.1 is necessary. There is, quite clearly, a link between morphosyntax and the classification scheme given. At times, a given word is classified partly because of its part of speech, e.g. when a word is acting as an adverbial booster it receives one label, when the same word form is acting as a premodifying adjectival intensifier it receives another label. One cannot, however, simply replace the labels with part-of-speech categories. For example, Curse, Dest, Gen and Literal are all examples where, for the word FUCK, the word is most likely to be a verb, as shown in the examples given in the table for this word. Yet in each case the use of the words, in functional terms, clearly differs. In Curse there is a clear insult intended, with a very clear target for the word. With Dest, while once again the intention to some degree is to insult,

‘So you recorded swearing’

27

there is also an imperative involved, typically with a demand being made that the target go away. In a Gen utterance, FUCK is used

Table 2.1 The categorisation of bad language Code

Description

PredNeg

Predicative negative adjective: ‘the film is shit’

AdvB

Adverbial booster: ‘Fucking marvellous’ ‘Fucking awful’

Curse

Cursing expletive: ‘Fuck You!/Me!/Him!/It!’

Dest

Destinational usage: ‘Fuck off!’ ‘He fucked off’

EmphAdv Emphatic adverb/adjective: ‘He fucking did it’ ‘in the fucking car’ Figurtv

Figurative extension of literal meaning: ‘to fuck about’

Gen

General expletive ‘(Oh) Fuck!’

Idiom

Idiomatic ‘set phrase’: ‘fuck all’ ‘give a fuck’

Literal

Literal usage denoting taboo referent: ‘We fucked’

Image

Imagery based on literal meaning: ‘kick shit out of’

PremNeg

Premodifying intensifying negative adjective: ‘the fucking idiot’

Pron

‘Pronominal’ form with undefined referent: ‘got shit to do’

Personal

Personal insult referring to defined entity: ‘You fuck!’/‘That fuck’

Reclaimed ‘Reclaimed’ usage—no negative intent, e.g. Niggers/Niggaz as used by African American rappers Oath

Religious oath used for emphasis: ‘by God’

Unc

Unclassifiable due to insufficient context

Table 2.2 Categories of annotation Field Feature marked

Possible values

1

Gender of speaker

M=male, F=female, X=unknown

2

Social class of speaker

As per social class categories of BNC (see Aston and Burnard 1998)

3

Age of speaker

As per age categories of BNC (see Aston and Burnard 1998)

4

Category of insult

As per Table 2.1

5

Gender of hearer

As per gender of speaker

6

Person of target

1=first person, 2=second person, 3=third person, X=unknown

7

Metalinguistic usage

0=no, 1=yes

8

Animacy of target

+=animate, −=non-animate, X=unknown

Swearing in English

28

9

Gender of target

As per gender of speaker

10

Number of target

1=singular, 2=plural, X=unknown

11

Quotation

Q=quotation, N=non-quotation, X=unknown

as an expression of general anger, annoyance or frustration. In the case of Literal, there is no clear intention to insult, merely an intent to describe an act of coitus. Parts of speech are clearly important to the categorisation scheme, but the scheme itself is not simply a relabelling of parts of speech. Also, and interestingly, just because a particular word covered in the LCA has a part of speech connected with a category does not mean to say that the word will appear in that category. For example, consider SHIT and FUCK. Both may clearly be nouns (the fuck I had last night was marvellous, I have just had a shit) or verbs (I was shitting when the cat came in, I fucked him). Morphosyntactically the words are similar. However, their distribution across the bad language categories differs—SHIT occurs only in categories PredNeg, AdvB, Figurtv, Gen, Idiom, Literal, Image, PremNeg, Pron and Personal. FUCK occurs in a larger number of categories—AdvB, Curse, Dest, EmphAdv, Figurtv, Gen, Idiom, Literal, PremNeg, Pron, Personal. While morphosyntactically similar, the words respond to these categories in different ways— FUCK has a much broader set of functions than SHIT. Yet there are some categories in which SHIT can be placed which do not apply to FUCK (e.g. PredNeg) and some categories FUCK can be placed in which do not apply to SHIT (e.g. Dest). It should also be noted that the distribution of examples across these categories is different—for example, EmphAdv accounts for most of the occurrences of FUCK in the LCA (55 per cent), while for SHIT most of its occurrences (44.6 per cent) occur in category Gen. BLWs do not act the same just because they share similar parts of speech. The range of classifications the word may express can differ, and their affinities for different categories in quantitative terms may differ radically, even where two words can appear in the same category. Given that the categories seem at least to display discriminating power, I will not seek to justify them further here. Before proceeding to explore the relationship of bad language to distinction, I will note once more that my focus here will be on the spoken language. Bad language in written English is still the subject of censorship. Hence, for any feature we observe—or fail to observe—in a written corpus, it is difficult to know whether it is an artefact of censorship or not. While there are limitations placed on speech also, these tend to be social rather than legal in nature, and hence the study of bad language in a corpus of spontaneous spoken English, such as that in the BNC, provides a much more secure basis for the study of bad language, as opposed to language which is subject to overt censorship.

Sex One might imagine that males use bad language more than females. Indeed it has been suggested in the past—especially in research in the 1970s—that swearing was a behaviour engaged in more frequently by males than by females,54 though recent research has retreated from that position to suggest that frequency for both sexes may vary

‘So you recorded swearing’

29

markedly depending on context and the gender of the hearer/hearers.55 However, it is still, in my opinion, a widely held folk belief in Britain that men swear more often than women. This is not the case. When all of the words in the LCA are considered, it is equally likely that bad language will be used by a male as by a female.56 A possible means of invalidating this finding, and to prove that males do indeed use BLWs more than females, would be to discover that a very frequent BLW used almost exclusively by males had been excluded from the LCA. This is clearly not the case. Another possible explanation is that a great number of low-frequency BLWs exclusively used by males had been omitted. Given the wide range of words included in the LCA, this also seems unlikely. A more likely explanation begins to emerge if we look at the distribution of BLWs between males and females. If we compare the BLW word forms used by males and females, we discover that there are a set of words significantly overused by males and a set of words significantly overused by females. If we look only at those words where a highly significant difference in use by males and females occurs (i.e. where there is a one-in-one-hundred chance or less that the result we have observed is attributable to chance) then 15 words emerge as being those which distinguish male and female swearing—fucking, fuck, jesus, cunt and fucker are, in descending order of significance, more typical of males; god, bloody, pig, hell, bugger, bitch, pissed, arsed, shit57 and pissy are, once again in descending order of significance, more typical of females. The results are shown in Table 2.3.58 So, while BLWs as a set may not differentiate males from females, the frequency of use of individual BLWs clearly does mark males and females apart. The words themselves suggest, to my intuition, another way in which males and females may differ. May it be the case that males have a preference for ‘stronger’ word forms while females have a preference for ‘weaker’ word

Table 2.3 Words preferred by males and females in the BNC ranked by LL value Word

god

Frequency of use by females per 1,000,000 words 459.38

Frequency of use by males per 1,000,000 words

Overuse by

Log-likelihood (LL) value

172.33 Females

549.09

284.10 Males

350.83

fucking

99.77

bloody

526.71

277.80 Females

314.15

fuck

32.75

68.28 Males

48.98

Pig

11.32

1.42 Females

36.55

hell

146.29

114.21 Females

15.69

bugger

39.48

25.00 Females

13.09

bitch

17.14

8.54 Females

11.82

pissed

24.18

13.82 Females

11.45

jesus

9.79

18.70 Males

10.88

Swearing in English

30

arsed

2.45

0.20 Females

9.44

cunt

5.51

11.18 Males

7.54

fucker

0.61

3.25 Males

7.41

shit

80.19

63.81 Females

7.38

pissy

1.22

0.00 Females

7.35

forms, i.e. those less likely to cause offence? It seems to me that the male marked words are more offensive, more potent, than the female marked words. However, in order to explore this hypothesis I need something more than my intuitions—I need access to something like a Richter scale for BLWs. It was beyond the scope of my work to conduct a large-scale survey to determine the strength of particular BLWs. Fortunately there was no need to do so, as such reviews have been commissioned by various media watchdogs in the UK. I have combined the results of two such surveys in order to provide a five-part scale of offence with which to classify the use of BLWs.59 The scale itself is borrowed from one of the sources that have contributed to its construction, the British Board of Film Classification. The scale is shown in Table 2.4. Using this scale it is possible to revisit Table 2.3 to explore the relationship between the strength of the words and such matters as gendered direction and speaker sex. In Table 2.560 each word is followed by a number in parentheses which indicate its position in the LL-score ranked table shown in Table 2.3. So, BLWs are a marker of distinction between males and females, but the distinction is marked quantitatively with a small set of word forms and is more generally marked qualitatively, with males drawing more typically from a stronger set of words than females. One further way in which male and female BLW use may differ is with reference to the types of BLWs discussed in the previous section. Are there categories of BLW use that are more markedly male or female? In order to explore this I contrasted the use

Table 2.4 A scale of offence Categorisation Words in the category Very mild

bird, bloody, crap, damn, god, hell, hussy, idiot, pig, pillock, sod, son-of-a-bitch, tart

Mild

arse, balls, bitch, bugger, christ, cow, dickhead, git, jesus, jew, moron, pissed off, screw, shit, slag, slut, sod, tit, tits, tosser

Moderate

arsehole, bastard, bollocks, gay, nigger, piss, paki, poofter, prick, shag, spastic, twat, wanker, whore

Strong

fuck

Very strong

cunt, motherfucker

‘So you recorded swearing’

31

Table 2.5 Table 2.3 revisited—BLWs typical of males and females mapped onto the scale of offence Shock

Male

Female

Very mild Mild

god (1), bloody (2), pig (3), hell (4) jesus (3)

bugger (5), bitch (6), pissed (7), arsed (8), shit (9)

Moderate Strong

fucking (1), fuck (2), fucker (5)

Very strong cunt (4)

of the different types of BLWs by males and females. Table 2.6 shows those categories which are overused by either males or females. It is interesting to note that Gen, a category which is not associated with abuse, is a more typically female than male category. Premodifying intensification and literal usage are also more typical of females than males. Again, neither is linked directly to personal abuse. The typically male usages are also, interestingly, not linked to abuse as such—they are both associated with intensification. So in terms of categories of swearing, it appears that only two styles of intensification—AdvB and EmphAdv—are markedly male. Note that the explanation for this may be linked to the male marked words. Fucking is the word which is most clearly male marked in the data (see Table 2.3). This word is only used by females 226 times in the LCA, 19 times as an AdvB and 154 times as an EmphAdv. By contrast the word is used 982 times by males, 107 times as an AdvB and 683 times as an EmphAdv. Given that there are in total 317 examples of BLWs in category AdvB and 1,953 examples of EmphAdv in the corpus, it is reasonable to say that the word fucking is a strongly male marked word, used mainly to

Table 2.6 Categories of BLW use more typical of males and females ranked by LL value Type of word Gen

Frequency of use by Frequency of use by males females

1,250 Females

100.09

1,131

822 Males

49.10

AdvB

202

115 Males

24.19

PremNeg

413

517 Females

11.65

58

90 Females

6.97

EmphAdv

Idiom

799

Overuse Log-likelihood (LL) by value

produce AdvB and EmphAdv effects. Given the ratio of all examples of AdvB and EmphAdv to those produced using the word fucking by males, AdvB and EmphAdv in turn become typical of male speech.

Swearing in English

32

The PremNeg category also bears some discussion, however, as it is a category within which fucking can and does occur. Might it be the case that the word fucking is a favourite form of female intensification and females reserve their use of the word fucking for the PremNeg form of intensification? The answer to this question is no—while males will often use fucking to realise PremNeg, females shun the word in the PremNeg category in favour of bloody. In the LCA, males use fucking as a PremNeg 158 times, in contrast to females who use it only 48 times. Yet females use bloody 422 times as a PremNeg, as opposed to males who use it only 126 times. Bloody is the word most typical of female intensification. Interestingly, the word does appear as both an AdvB and an EmphAdv, but is not used by males and females differently in a way that is statistically significant (AdvB 75 and 76 times respectively, EmphAdv 426 and 650 times respectively). It is only in the PremNeg category that a statistically significant difference emerges. So bloody as a typically female word (see Table 2.3) and PremNeg as a typically female intensification strategy seem to be choices which are complementary to those of males, who prefer fucking and AdvB/EmphAdv type intensification. As such, intensification is an area in which there is a marked distinction between males and females both in terms of lexical choice and BLW category choice. Yet speaker sex is not the only gender variable in the LCA. How do speakers respond to speakers of the same, or different, sex? Do males act, as one would imagine that gentlemen may, to avoid BLW use in the presence of ladies? Do they refrain from directing BLWs at ladies? With regard to women, do we find that women are less likely to use BLWs in the presence of other women? Do they also prefer not to direct BLWs at other women? The LCA can shed some light on these questions, though not really on the hearer-related questions—as noted, deriving hearer gender from the BNC is difficult. So answering the questions about whether BLW use will take place in the presence of people of a certain sex is impossible with BNC derived data. It is possible to study the gender of the direction of BLW use, however. If one limits one’s questions to examples where there is a clear direction to the BLW use, one is able to explore whether males direct BLWs at males more often than females, whether females direct BLWs more often at males than females, etc. When one explores these issues in the LCA, one observation can be made immediately: BLWs are directed at males more often than females.61 While this shows us something about language directed at males, it shows us little about the patterns of interactions behind that general pattern. Who is it that is swearing at males more than females? Other males? Females? In order to explore these questions I constructed the matrix shown in Table 2.7. This table shows how often one gender directs BLWs at either gender in the LCA. All of the pairwise comparisons one may want to make in Table 2.7 are significant. Males direct BLWs at a male target far more often than they do at a female. Exactly the reverse is true of women—they are more likely to direct BLWs at other women. At this point the question of hearer becomes relevant again. Is it the case that women use BLWs more in the presence of other women, but suppress them in male company so as to appear more ladylike’? While this is certainly a possibility, the inability to annotate hearer information in the corpus precludes the exploration of this hypothesis here. However, what can be seen very clearly are preferences for an intragender direction of BLWs for both sexes. This result only holds if we treat all BLWs as identical, however. If we assume that some BLWs may prefer intragender direction, while others prefer intergender

‘So you recorded swearing’

33

direction, and study gendered direction at the word level, we may discover not only that certain words do prefer intergender direction, but also that certain words are exclusively directed at one sex rather than the other. Tables 2.8 and 2.9 explore these questions. As can be seen from Tables 2.8 and 2.9, for both sexes there is some degree of differentiation of the use of BLWs for targets of different sexes with, for example, the words cow, bitch, bloody, fucking, slag, tart, tit, tits and whore showing a pronounced bias for being directed at females by females and god, bastard, gay, christ, git and cunt being words which, when used by females, show a pronounced bias towards being directed at males. The words themselves seem to display some evidence of being directed quite differently by males and females. For example, the word cunt is directed exclusively at males by females. It is a pure intergender BLW for females. This is not true

Table 2.7 Patterns of male/female-directed BLW use Male directed

Female directed

LL score

Male speaker

702

156

375.82

Female speaker

392

497

12.43

89.06

187.2

LL score

Table 2.8 Words more likely to be directed by females at either males or females ranked by LL score Word

Female targets

Male targets

Preferred target

cow

25

god

16

55 Male

22.66

2

24 Male

21.94

bastard bitch bloody gay fucking

0 Female

LL score 34.66

26

3 Female

20.91

153

85 Female

19.70

12 Male

16.64

24 Female

14.53

0 58

christ

0

10 Male

13.86

git

0

9 Male

12.48

slag

9

0 Female

12.48

cunt

0

8 Male

11.09

tart

8

0 Female

11.09

tit

6

0 Female

8.32

Swearing in English

34

tits

6

0 Females

8.32

whore

6

0 Female

8.32

Table 2.9 Words more likely to be directed by males at either males or females ranked by LL score Word fucking

Female target

Male target

Preferred target

LL score

37

254 Males

181.71

1

35 Males

40.77

25

89 Males

38.11

fuck

1

33 Males

38.11

pissed

0

27 Males

37.43

bastard

2

30 Males

29.40

wanker

0

13 Males

18.02

cunt

1

15 Males

14.70

arse

5

25 Males

14.56

fucked

0

10 Males

13.86

shit

3

19 Males

12.97

piss

2

16 Males

12.40

git

1

13 Males

12.20

bird

7

0 Females

9.70

christ

0

7 Males

9.70

gay

0

7 Males

9.70

sod

1

10 Males

8.55

arsehole

0

6 Males

8.32

prick

0

6 Males

8.32

cow

9

1 Females

7.36

idiot

1

9 Males

7.36

god bloody

for males, who, while showing a strong preference for directing the word at males, do also direct it at females. Cow is a pure intragender word for females; this does not appear to be the case for males. So the question of whether or not certain words may be directed at one sex or another may very well depend on the speaker. However, there is doubtless another set of words in the LCA which have purely male and female targets, irrespective of the gender of the speaker. These are listed in Table 2.10.62 The figures in parentheses

‘So you recorded swearing’

35

indicate the number of occasions in the LCA in which the word is directed at either a male or female, as appropriate. In very few instances in the table are the figures large enough to allow one to generate a claim of statistical significance. Nonetheless, it is interesting to note that there do appear to be some words which are exclusively male directed and exclusively female directed irrespective of the gender of the speaker. This is certainly the case for words like whore, for example, which is directed at females by females six times and by males at females twice. Similarly, gay is directed exclusively at males, seven times by men, 12 times by women.63 Note that the gay example is an interesting one—it has taken on the role of being a BLW applied exclusively to males, while this need not be the case, as the term can cover males and females. Also, it has retained that gender-exclusive direction even though other words which one might assume would exhibit such a gender exclusive direction, such as bitch, do not exhibit such exclusivity.64 So, it seems to be the case that there is a set of BLWs which are used with an exclusive gender direction by male speakers but not female speakers (e.g. arsehole).65 Similarly there is a set of BLWs used with an exclusive gender direction by female speakers but not males (e.g. cow). There is a third set of words which are exclusively directed at a particular gender by a speaker irrespective of the sex of the speaker (e.g. gay). Over and above that, there are words which show a strong gender targeting/production preference for/by males or females but which are not gender exclusive (e.g. bloody tends to be targeted at females by females, bastard tends to be directed at males by males). Given that there are at least three types of words where some degree of gender exclusivity in direction applies, an obvious question one should ask is whether there is anything that typifies the words targeted at either gender.

Table 2.10 BLWs directed solely at males and females ranked by frequency of usage Female-only direction slag (12), bird (10), tart (10), tits (9), whore (8), tit (7), birds (3), hussy (3), whores (3), bitches (2), slagged (2), tarty (2) Male-only direction gay (19), christ (17), wanker (16), prick (7), pillock (4), poofter (4), fucker (3), moron (3), pissing (3), pissy (2)

Given that speaker sex is an important variable for two of these sets of words, and that sex is an important variable in dictating the direction of the words, gender seems, once more, an interesting variable to explore. May it be that, in some qualitative sense, the BLWs in the three sets vary? One obvious way in which they may vary is in the strength of the words. Are males and females as harsh in their use of gender exclusive/biased BLWs? Do males use weaker BLWs when directing them towards females rather than males? A host of questions may be related to the question of gender, direction of BLW, speaker and strength of the language employed.

Swearing in English

36

Tables 2.11 and 2.12 also suggest that intragender BLW use is more frequent generally than intergender BLW use. They also make it fairly obvious, once more, that stronger BLWs are more often directed at males than females. May it simply be the case that, when it comes to gender exclusive/biased BLWs, this is true, but for all of the other BLWs it is not, meaning that, overall, it is not the case that there is any differentiation in the strength of BLWs directed at males and females? This question was explored to some extent in Table 2.5, but is worth revisiting. In order to explore this question further I assigned a value to each of the BLWs appearing in Table 2.10, with the words in the mildest category assigned an offence strength of one, with the offence strength increasing by category to a

Table 2.11 Table 2.8 revisited—BLWs typical of females used either of males or females mapped onto the scale of offence Shock

Male target

Female target

Very mild

god (55)

bloody (153), tart (8)

Mild

christ (10), git (9)

cow (25), slag (9), tit (6), tits (6)

Moderate

bastard (24), gay (12)

bitch (26), whore (6)

Strong Very strong

fucking (58) cunt (8)

Table 2.12 Table 2.9 revisited—BLWs typical of males used either of males or females mapped onto the scale of offence Shock

Male target

Female target

Very mild

god (35), bloody (89), idiot (9)

bird (7)

Mild

pissed off (27), arse (25), shit (19), git (13), christ (7), sod (10)

cow (9)

Moderate

bastard (30), wanker (13), piss off (16), gay (7), arsehole (6), prick (6)

Strong

fucking (254), fuck (33), fucked (10)

Very strong

cunt (15)

maximum of five for a word such as cunt.66 Then, the number of times each word was directed at males was multiplied by the strength of the word. This was repeated for all words and the results added together and divided by the total number of words under consideration to yield an average strength of a BLW directed at a male by a male. The process was then repeated with all combinations of male/female direction. The result is that females target females with an average BLW strength of 1.956 and males with an

‘So you recorded swearing’

37

average BLW strength of 2.042. Males target females with an average BLW strength of 1.562 and males with a BLW strength of 2.779.67 So, while the frequency of BLW is higher in intragender BLW use, in terms of strength males have stronger BLWs targeted at them by speakers of both sexes. While females may direct BLWs at females more frequently, when they direct them at males, they select stronger BLWs than when they are directing them at females. Males, on the other hand, both direct BLWs less frequently at females than males and use weaker BLWs when directing BLWs at females rather than males. May the pattern change if we look at BLWs independent of target, i.e. include nontargeted BLWs (e.g. Gen types)? The average strength of a BLW spoken by a female, irrespective of target, is 1.48. For males it is 2.23.68 It is apparent that not only do females generally use ‘weaker’ BLWs, they also have weaker BLWs directed at them. Males, on the other hand, use stronger BLWs on average and have stronger BLWs directed at them, though when directing BLWs at females, as we have seen, they use weaker BLWs than females would.69 Before leaving the issue of strength, may it also be the case that the affinities of the different types of BLW use to the different sexes is related to the association of these categories with stronger or weaker forms of BLW use? This is indeed the case. Table 2.13 gives a rank ordered scale of the average strength used in each BLW category. This table deserves some discussion. One possible response to the finding is to note, for example, that most examples of the word FUCK occur in the Curse and Dest categories. As FUCK has a high strength, the category has a high strength. However, this is overlooking the versatility of FUCK and other BLWs. There are a wide range of words which can be used to realise a Curse-type BLW expression, or a Dest type. For example, bugger can also be a Curse. Yet there are almost twice as many (49 v. 29 examples) of fuck as a Curse compared to bugger as a Curse. There is a clear choice by speakers to select the stronger word when producing an expression which would fall into this category. Similarly, for category Personal, both bugger and fuck are possible choices when realising a Personal type utterance. Yet in this category bugger is more frequent than fuck (194 v. 8 examples). There is evidence that some categories select the stronger words while others select weaker words. A good example of this is category Gen—the BLWs (in descending order of frequency, with the frequency of the appearance of this word following the word in parentheses) in the Gen category are as follows:

Table 2.13 Average strength of BLWs in each category Category

Frequency in LCA

Average strength of word in the category

Dest

110

3.37

Curse

105

3.10

22

3.00

Personal

756

2.48

Literal

144

2.41

1,948

2.29

Reclaimed

EmphAdv

Swearing in English

38

AdvB

312

2.22

Figurtv

384

2.01

Image

108

1.94

36

1.79

PremNeg

828

1.69

Idiom

453

1.52

PredNeg

101

1.37

Gen

2,048

1.29

Oath

53

1.01

Pron

god (1,288), hell (281), shit (159), christ (109), jesus (92), fuck (70), damn (16), bugger (11), crap (8), gods (2), piss (2), shite (2), bloody (1). Other than fuck, there is not a single word in the Gen category with a strength higher than 2.70 Gen, as a category, seems to select milder words. Given that many of the words which mark out differences between males and females are linked to these categories (e.g. god as a weak word preferred by females, strongly associated with Gen-type BLW use) a fuller picture emerges of the differences between male and female BLW use, with females using weaker words in consequence of using weaker categories of BLW use (or, conversely, choosing weaker categories of BLW use because they wish to use the weaker words). Sex as a variable will be revisited throughout this chapter. This initial investigation of gender-based differences is interesting in that it demonstrates the relative subtlety of the differences in BLWs between males and females. It also demonstrates the utility of using the LCA for such an investigation. Let us now explore another factor: age.

Age Again, the literature would lead one to believe that age is an important variable in the use of BLWs. For example, Cheshire (1982:101) claimed that swearing has a particular value for teenagers, as it is a ‘major symbol of vernacular identity’ for this age group. Hence one would expect to find more swearing in that group. This is indeed the case. When one looks at all of the BLWs in the LCA, there is a positive correlation between age and the production of BLWs, though the pattern is not quite as straightforward as it at first appears. In rank terms, using the same age intervals as those in the spoken section of the British National Corpus, the profile of BLW use shown in Figure 2.1 emerges. This figure shows that BLW use increases into the age range U25 and thereafter generally steadily declines.71 This result holds for both males and females. This graph certainly lends some support to the hypothesis that adolescents are more likely to use BLWs, perhaps for the reasons that Cheshire outlined. While it would be nice to have a corpus in which actual ages, rather than ages within a band, were recorded so that the U15 and U25 categories could be further broken down, it is far from implausible to suggest that there could well be a hidden peak in BLW use in this graph in the adolescent

‘So you recorded swearing’

39

age range. This peak may be responsible for the high volume of U15 data and may largely account for the peak in the U25 data. While this must remain speculative, given that the hypothesis is in line with the predictions and findings of others, I would venture to say that this hidden peak is almost certainly present in this data. Hidden peaks aside, some questions may still be asked of these results. May the results be a side-effect of the omission of one or more very frequent BLWs used by the older speakers? There is no evidence for this in my exploration of the BNC and the LCA. The only possible grounds on which such a result could rest relates, in my view, to such matters as euphemisms or very, very mild realisations of BLW categories (e.g. dear in ‘Oh dear!’). It may be the case that the older speakers produce more very weak BLW utterances, avoiding the direct use of BLWs. So, for example, they may say ‘oh fudge’ or ‘oh flip’ rather than ‘oh fuck’. As my study does not cover euphemism, I have not recorded or explored such cases. To do so would require a very labour intensive, manual exploration of the BNC. However, in

Figure 2.1 Frequency of BLWs per million words in groups of different ages. such an exploration, the categories developed in this study may be of help—most of them are associated with fairly fixed frameworks which help to generate an utterance of a certain category. So, for example, ‘Oh’ followed by a BLW typically generates a Gen.72 If we find other words other than a BLW in such a context, we also have the potential for a Gen, e.g. ‘oh flip’, ‘oh sugar’, ‘oh fiddlesticks’. In order to check the potential impact of such omissions on my results, I explored the pattern ‘Oh+any word+!’ in the spoken BNC.73 The results do not indicate that large numbers of Gen types were missed by the LCA. Those that were missed for this pattern would not have boosted the frequency of Gen types for the older speakers. In fact, it would have done quite the reverse. The search gave rise to five Gen types for males under 15 formed with the words, sugar, dear, boy, blimey and golly. For females, two new Gen types were spotted, both produced by under-

Swearing in English

40

15-year-olds and both formed with dear. From this, admittedly limited study, there is no evidence that a vast swathe of BLW-like events are not represented in the LCA, and there is no evidence that this leads to the frequency of BLWs produced by older speakers being under-reported, though there is the scantiest indication that BLW-like utterances by the under-15s may be under-reported. The next obvious question one may ask of the data is will the order reported here change if we take the strength of the BLWs produced by speakers of different ages into account? The answer is no; when one calculates the average strength of a BLW produced by speakers in the different age groups, the results powerfully reinforce the negative correlation of age with swearing. The results are: U15–2.24, U25–2.31, U35–2.16, U45– 1.31, U60–1.55 and 60+ 1.17. While the correlation with age is, once again, not perfect, the correlation is nonetheless noticeable and generally negative. Given that there is a diachronic dimension to comparing speakers of different ages, one does need to consider whether one is observing language change, or whether the snapshot provided by the BNC is typical of BLW use in any era. This is a difficult question to address given the corpus resources available. If in the future an equivalent to the spoken BNC is produced, it may be possible to explore changing patterns of BLW use over time. As it stands, however, this is not possible at the moment. On the other hand, given that the literature on BLW use argues that the correlation observed here should indeed be observed, I will assume, until evidence to the contrary presents itself, that what is observed here is what researchers have expected to see for some time—a correlation between age and BLW use, with BLW use declining as speakers become more conservative with age. Before leaving the discussion of age, it is worth considering whether the different age groups select from the BLW categories differently, and whether the particular age groups have particular preferences for certain BLW categories. With respect to the question of whether the age groups vary in their BLW category choice, the answer is yes.74 The results of an exploration of which age group is the most, and which is the least, frequent user of each type of BLW use are given in Table 2.14. The results for the 60+ category is worth discussing—the evidence is that, pretty much across the board, this group uses BLWs least and avoids the largest number of BLWs as a consequence. In the case of Reclaimed, although this group uses the largest proportion of all recorded examples of this category other than the U15 age group, this result should be treated with caution as Reclaimed is the least frequent category in the LCA (see Table 2.13)—there are only five recorded examples of the 60+ group using a Reclaimed in the corpus. With reference to Oath, this is the weakest category of BLW use. Hence its relatively more frequent use by the U60 (and 60+) group is explicable purely in terms of them using these weaker BLW forms. Note that the avoidance of this category by the U25 group is explicable on the same grounds, as is the attraction of the U15 groups to stronger BLW categories on the whole. The U15 and U25 groups are attracted to much stronger BLWs (notably those forming Dest and Personal types). They correspondingly avoid the weaker BLW categories, leading to their avoidance of Oath types. Table 2.14 in conjunction with Table 2.13 clearly suggests a relationship between BLW category strength and age.

‘So you recorded swearing’

41

Within the age groups, this is reinforced slightly. In Table 2.15, the top four ranking categories for each age group are given. Following each type in the table is a figure indicating its rank in the strength of BLW table (with 1 being strongest, see Table 2.13). For each age group the picture is remarkably similar—they use EmphAdv, Gen, Personal and PremNeg types most frequently, except the

Table 2.14 The most frequent and least frequent users of particular BLW categories, categories ranked by strength from highest to lowest Category

Highest use

Lowest use

Dest

U25

60+

Curse

U35/U60

60+

Reclaimed

U15

U35/U45

Personal

U15

60+

Literal

U15

60+

EmphAdv

U60

60+

AdvB

U35

60+

Figurtv

U25

60+

Image

U25

60+

Pron

U15

60+

PremNeg

U60

60+

Idiom

U60

60+

PredNeg

U15

60+

Gen

U25

60+

Oath

U60

U25

Table 2.15 The top-four BLW categories for each age group Rank Type U15

Type U25

Type U35

1

Gen (14)

Gen (14)

EmphAdv (6) Gen (14)

2

Personal (4)

EmphAdv (6) Gen (14)

3

EmphAdv (6) Personal (4)

4

PremNeg (11)

PremNeg (11)

Type U45

Type U60

Type 60+

EmphAdv (6) EmphAdv (6)

EmphAdv (6) Gen (14)

Gen (14)

Personal (4)

PremNeg (11)

PremNeg (11)

PremNeg (11)

PremNeg (11)

Personal (4)

Personal (4)

Idiom (12)

Swearing in English

42

60+ age group, where Personal does not feature in the top four, being usurped by the much weaker Idiom type. While the ordering of EmphAdv, Gen, Personal and PremNeg varies somewhat by age group, it is clearly in the 60+ age group that the major change occurs. However, looking across the table, one might claim again that age and strength interact—Personal, the strongest category in the table, declines steadily in the rank ordering as the speakers grow older, with the rank profile being 2, 3, 3, 4, 4 before the Personal type exits the top four in the 60+ column. So, with specific reference to the Personal type, we might suggest that, once again, Table 2.15 furnishes evidence that age and strength of BLW use is correlated. An obvious problem, however, is that within each age category are respondents who differ in a number of potentially important ways, e.g. sex and social class. While I have already made some observations about sex, I have yet to consider social class. In the next section I will, therefore, consider social class as a variable before returning to the question of how age, sex and social class may interact.

Social class What of social class and BLW use? Does BLW use simply decline as we look higher up the social hierarchy? When tested, the differences in the use of BLWs by different social classes is indeed significantly different.75 Does this clear distinction between the classes shown in Figure 2.2 also hold for strength of BLW use—do the lower social classes select stronger BLWs, while the higher social classes select weaker ones? When one calculates the average strength of BLW for each social class, a different picture

Figure 2.2 Frequency of BLWs per million words of speech produced by different social classes.

‘So you recorded swearing’

43

emerges: ABs use slightly stronger words than C1s. The average strength of BLW used by each social class is as follows: AB–1.81, C1–1.76, C2–2.16 and DE 2.47. In short, the rank order changes slightly from that produced by frequency alone and it no longer aligns itself neatly with the social class hierarchy, with the rank order being DE >C2>AB>C1. While AB speakers use BLWs less frequently than C1 speakers, on average they use BLWs with a greater strength than C1s. This could be evidence of hypercorrection76 by C1 speakers: in attempting to copy the linguistic habits of the AB social class, the lowermiddle-class speakers exaggerate what they view to be a feature of AB speech, i.e. the avoidance of strong BLWs. However, they do it to such a degree that they avoid strong BLWs more than the ABs do. While I will not discuss this hypothesis further here, it certainly provides an explanation for many of the differences between C1 and AB speech presented in the rest of this chapter. What of the type of BLW use undertaken? In terms of proportionate contribution to each category of BLW use, social class DE is attracted fairly uniformly to all forms of BLW usage except for PredNeg, Literal and Pron type BLW use.77 Speakers of social class AB are the most frequent users of these categories of BLWs. At the other end of the spectrum, as we would expect from the results presented in this section so far, C1 speakers are the opposite of DEs: they are infrequent users of almost every category of BLW use. There are, however, two exceptions. C2s shun PredNeg types more than C1s, and C2s shun Gen types more than C1s. The result with the PredNeg types is particularly interesting. With this type of BLW we see a marked change of behaviour for the C1s, which may be caused by the preference of ABs for that category, i.e. it may be the case that this category of BLW use is attractive to the C1s because it is strongly associated with a higher social grouping. However, there is another, possibly complementary, explanation for the use of PredNegs and Gens by the C1s—they are two BLW categories associated with weaker BLWs, as explored in pages 42–43 of this chapter. The use of As and Gs by the C1s may simply be a reflection of their use of weaker BLWs selected by weaker BLW categories. Similarly, when one considers the behaviour of the ABs, their use of As may be a reflection of the relatively weak nature of this category. The same is clearly less true for Pron and apparently not true at all of Literal-type BLW use. However, with reference to Literal usage, a question which cannot be explored here, as it was not explored in the studies on which the strength of the BLW scale was based, is whether when BLWs have a literal meaning they are less offensive than when they are being used, for example, to form a personal insult. For example, is the use of FUCK in ‘I fucked her’ less offensive than its use in ‘I cannot stand that fucking Dean’? If this were true, then there would be a much more persuasive case for the usurpation of the dominance of the DEs in certain BLW use categories by ABs being related to the strength of the word involved. However, on the basis of current evidence this must remain speculation. What if, rather than looking across the classes, one focuses on each class to see which categories of BLW each class favours and which it shuns? Broadly speaking, all of the social classes use BLWs of different categories in broadly similar proportions. There is relatively little variation as one moves from one social class to another with regard to how frequently they choose each category to express a BLW. Yet there are differences. If we focus on the top four categories of BLW use per social class this becomes apparent. For DE and C2 speakers they are identical: EmphAdv>Gen>PremNeg>Personal. For AB

Swearing in English

44

and C1 speakers, the same four categories also represent the top four, but the order differs: for AB and C1 speakers, the top two categories are Gen > EmphAdv. For ABs these are followed by Personal> PremNeg, while for C1s the patterning of the C2/DE group reasserts itself to produce PremNeg > Personal. In short, C2 and DE are identical, C1 is very similar to C2/DE, and for ABs there is not a single BLW type ranked the same as for C2s and DEs in the top four. As a further argument for social assimilation, the case is persuasive. AB is different from C2/DE speech, yet C1 speech positions itself between AB and C2/DE speech, at least with reference to BLW use. Looking at the top four most frequent BLWs for each class reveals a very similar pattern across the classes once again. The top four BLWs (lemmatised, in rank order) are: for ABs GOD>BLOODY>FUCK>HELL, for C1s GOD >BLOODY>FUCK>HELL, for C2s BLOODY>FUCK>GOD>HELL and for DEs FUCK>BLOODY>GOD>HELL. Once again, the ABs and C1s are identical, with frequency being roughly in inverse proportion to strength of BLW. Yet as we move through the social classes, this correlation reverses itself and the top four changes so that the rough correlation between frequency and strength of BLW become inverted, i.e. the more frequent the word, the higher its strength. Overall, the patterning, and the changes to the patterning as one moves through the social classes, reinforce the findings based on category of BLW use, though in this case the similarity of C1 to AB speech is much more compelling. As well as BLW type interacting with class, gender may also interact with social class and BLW use. One way in which it may do so relates to the likelihood of a BLW being directed at a particular gender. It is plausible that the BLWs directed at one gender or another by speakers may vary by social class. Intuitively, one might expect that, as one goes up the scale of social class, the likelihood of BLWs being targeted at females may decrease, as this language would increasingly be viewed as something which it was not fitting to use in the presence of or with reference to a woman. When this intuition is explored using the LCA it is proved to be partly wrong: there is a difference in the targeting of males and females with BLWs, but it varies in ways which are more akin to the overall pattern of variation for BLWs and class. When we look at the rank ordering of BLWs targeted at males, the rank ordering is thus (the occurrences per million tokens directed at males by that particular social class are given in parentheses): DE (2,567.36)>C2 (852.17)>AB (671.85)>C1 (284.54). The pattern is exactly the same as the overall pattern for the selection of BLWs of a greater strength by the different social classes. More surprising is the result of the exploration of female directed BLW usage, the rank ordering is: DE (1,260.99)>AB (499.69)>C2(195.54)>C1 (106.11). When these results are compared pair-wise within each social class, it is still the case that each social class differs in its approach to BLW use directed towards males and females, directing significantly fewer BLWs at females than males.78 However, the rank ordering clearly shows us that for the AB class this trend is less pronounced to the degree that, when considered as a proportion of all BLWs targeted at either sex, ABs are the third largest users of BLWs targeted at males, but the second largest users of BLWs targeted at females. So while class relates to BLW use in ways in which we might expect (frequency of usage being inverse to height of social class) there is evidence to suggest that class also interacts as a variable with BLW use in ways we would not expect, with the highest social class in the BNC, AB, sometimes bucking the trend the other social classes

‘So you recorded swearing’

45

conform to by using more and stronger BLWs directed more indiscriminately at both males and females. Yet what of the variables in combination? May age, sex and social class interact? This question is explored in the following section. Combining factors79 At this stage in the study, it would be useful to look at how the age, sex and social class variables interact in the BNC. This, however, is not possible with all of the BLWs studied so far. The reason for this relates to what happens to the available data in the BNC when we combine these factors: the corpus becomes very unbalanced. When we study one factor alone in the BNC, there is plenty of data in the corpus to populate the different categories. However, when we combine the factors, some categories have scarcely any examples in them at all. Take the example of the speakers of a specific age, sex and social class in the BNC who use very few BLWs (as recorded in the LCA): male AB speakers aged 25–34, male C1 speakers aged 15–24 and female DE speakers aged 0–14. The identification of these three groups as infrequent BLW users seems to challenge some of the general results presented in the previous sections—we appear to have found three groups who hardly use BLWs at all even though they are in age groups, and often in social classes, in which we would expect relatively frequent BLW usage.80 But in fact the answer to this conundrum is rather simple—there is hardly any data for those speakers in the BNC. Table 2.16 shows how many words are uttered by speakers from these groups in the spoken section of the BNC.81 The apparent avoidance of BLWs by these groups is illusory—there is simply a lack of data for these groups. Yet even where there is apparently

Table 2.16 The number of words spoken by three categories of speaker in the spoken BNC Group

Total words uttered by the group in the spoken BNC

Male AB speakers aged 25–34

2,259

Male C1 speakers aged 15–24

3,796

Female DE speakers aged 0–14

812

plentiful data, the number of groups created by combining factors together can spread the existing data so thinly that it ceases to be useful. Take the word bastard for example. There are 31 examples of the word in the LCA. However, when we combine age, sex and social class, the result is 48 categories (2×4×6). There are more categories than examples leading, not surprisingly, to 25 of the categories being empty when the examples of bastard are assigned to them. Hardly the basis for an illuminating investigation of the interaction of age, sex, social class and BLWs. While disappointing, this does serve to show that, even in a large corpus such as the BNC spoken section, data sparsity can be a very real problem.82 Nonetheless it is possible to look at some of the interactions of age, social class and sex with reference to a number of high-frequency words, namely bloody, fucking and shit.

Swearing in English

46

In order to do this the frequency counts of these words were subjected to a further statistical test: log-linear modelling. In the first phase of this test, the model tested to see whether the words varied within each variable, i.e. it sought to verify the findings presented so far. The model verified those findings, indicating that the differences for the different subcategories of age, sex and social class gave rise to significant differences of use by the different speakers in the subcategories.83 This result, while useful as it corroborated the findings presented so far, did not look at the interaction of the variables. The next stage of the test was to look at whether pairs of variables generated subcategories across which the use of these words varied significantly. When looking at age and class, gender and age, and gender and class, the model returned a significant result each time, indicating that there was an interaction between the variables which influenced the frequency of the use of the three words under investigation. Before considering what the nature of that relationship is, however, the issue of data sparsity becomes relevant once more. While gender and class together generate a relatively limited set of categories (i.e. number of categories of sex multiplied by the number of categories for class, 2×4=8) the number of categories increase when we consider gender and age (2×6=12) and becomes large when we consider age and class (4×6=24). While there is sufficient data to reliably test the interaction of gender and class and gender and age, the number of categories generated by age and class is too great—the results become unreliable.84 This means that, unfortunately, any results in this category can best be described as indicative only—they are certainly not something one would want to rely on. Unsurprisingly, therefore, even with these frequent words we are not able to study the interaction of all three variables as the number of categories (48) that this produces is far too great. It should be noted, though, that when this model is tested, the interaction of all three is significant, but the model is once again so sparsely populated that it would be wisest to discard the result. However, we can reliably test the interaction of gender and age and gender and class. As these have proved to be significant, the following subsections will explore the nature of these two relationships. Age and sex The log-linear model is very useful when identifying if a relationship exists between two variables. It does not, in itself, indicate very clearly exactly what the nature of the relationship is. To discover that, we need to return to the raw data and look at the pattern of interaction set up by the two variables. This is easily done—we simply need to look at the distribution of the uses of the three words across the eight categories generated by the interaction of age and sex. Table 2.17 shows the interaction between these two variables. What is apparent from the table is that the single behaviours—gender differentiation being apparent for certain BLWs, and BLW use peaking in the U25 age group and declining thereafter—generally hold in this table. There are exceptions to the general pattern, however, with most of the exceptions being related to the word bloody. In males, the use of bloody peaks later—in the U35 category—and its decline thereafter is far from marked, with the frequency rising again from the U60 category onwards. For females, there is an early peak in the U25 category, but once again a second

‘So you recorded swearing’

47

Table 2.17 The interaction of age and sex,, frequencies given as normalised counts per million words Word

U15

U25

U35

U45

U60

60+

Male bloody

87.85

154.82

416.94

305.73

362.50

372.75

fucking

111.64

903.46

875.43

33.56

191.24

6.55

shit

118.06

168.97

126.62

50.09

18.98

9.70

bloody

27.95

348.99

342.30

157.57

552.81

121.49

fucking

43.39

149.99

70.23

10.04

120.28

0.00

shit

37.64

208.64

7.74

9.16

26.96

1.19

426.53

1,934.88

1,839.26

566.16

1,272.77

511.69

Female

Total

peak, and in this case the highest peak of usage, occurs in the U60 category. In both cases, the peaks for bloody usage by females far exceed those for the males. The word fucking also gives rise to a double peak, for both males and females in the U25 and U60 categories, but in this case the male peaks exceed the female peaks. The word shit shows a double peak for females only, once again in the U25 and U60 categories. If there were a much larger data set in which one could view this twin peak patterning for bloody against the patterns achieved for other words, a number of hypotheses could be explored to explain the usage of the word. In the absence of that data set, I will permit myself a little intuitive speculation. Given that the general pattern of BLW use with age is a decline in usage beyond U25—a pattern attested here only with reference to the word shit used by males—how can the double peaks for bloody and fucking be explained? The higher peaks for bloody when used by females as opposed to males is easily explained— as shown on page 35, bloody is a BLW used more heavily by females than males, hence its relative over-use here is quite understandable. What is more difficult to explain is why the second peak occurs. My hypothesis would be that there is a narrowing in the BLW lexicon with age. While BLWs as a group may be used less frequently with age, the number of BLWs called on shrinks more swiftly so that, almost paradoxically, the selection of a specific BLW in the age range U60 may peak temporarily as the range of words used to realise a BLW utterance shrinks more markedly than the rate of decline of BLW usage. This gives rise to a series of local peaks in BLW usage for certain words in the U60 age range which are flattened as age increases and BLW use declines further, meaning that, even for the few BLW words still used by the speaker, their rate of BLW usage is now so low that the frequencies of the surviving BLWs are reducing again. To thoroughly test this hypothesis would, of course, require a much larger corpus than the one I have available. However, some persuasive evidence can be provided from the BNC if one maps the usage of different BLW word forms (types) against age. The results show

Swearing in English

48

a very familiar pattern—the number of different word forms related to BLW usage peaks in the age range U25 and declines thereafter. The results are shown in Table 2.18. Table 2.18, while in no way proving my hypothesis, certainly lends it support. There is a marked reduction in the number of types of BLWs used by the U60 age group, with a further marked reduction in the 60+ age group. Hence the hypothesis that the shrinking BLW lexicon of the U60s

Table 2.18 The number of different word forms used to realise BLW use by the different age groups in the LCA Word count

U15

U25

U35

U45

U60

60+

71

75

58

51

43

29

causes a peak in usage for some of their remaining BLWs is certainly plausible.85 Yet much more work is needed before that hypothesis can be stated with anything approaching certainty.86 Given that such work is well beyond the scope of a sole scholar, I leave this hypothesis to others to investigate further. Before leaving the discussion of age and sex, however, it is worth considering the claims made about age and sex by other researchers, using smaller data sets. One major problem with these studies, such as Eiskovits (1998), is that they have tended to treat words as a lumpen mass and have looked to see how gender relates to that mass of words. As shown in this chapter, and this section, words which might be lumped together, e.g. BLWs, may respond to factors such as gender and age very differently. Hence when Eiskovits studied non-standard features, the results may have over-generalised how gender and non-standard language relate to one another. For example, Eiskovits studies non-standard grammar and swearing.87 Her conclusion was that girls modify their speech in the direction of the standard as they moved out of adolescence while boys increased their use of nonstandard features such as swearing. However, as can be seen from the data in this section, this may not be true of certain word forms. Nor may it be true in general when attested language use, rather than reports of language use, which Eiskovits relied on for her comments on swearing, is considered. One might modify what researchers such as Eiskovits claim to say that, as males and females move out of adolescence, their use of BLWs becomes more gender differentiated, with frequency of use of these words generally declining over time, as all BLW usage declines. But if the findings presented in this chapter so far show anything, they certainly show that grouping words, while useful, should always be balanced by the study of words in isolation when considering how words and sociolinguistic variables interact. Sex and social class Table 2.19 shows the interaction between the variables sex and social class. A general trend is obvious from the table—female BLW use meets the expectations we have for speakers of different social classes more readily than male BLW use does. The rank of BLW use is in an inverse relationship with social class for females: the higher the social class, the lower the usage of BLWs. While the frequency profile of

‘So you recorded swearing’

49

BLW usage for females is AB